<?xml version="1.0" encoding="UTF-8"?>
<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9" xmlns:image="http://www.google.com/schemas/sitemap-image/1.1" xmlns:xhtml="http://www.w3.org/1999/xhtml" xmlns:video="http://www.google.com/schemas/sitemap-video/1.1">
  <url>
    <loc>https://www.kr.nota.ai/community</loc>
    <changefreq>daily</changefreq>
    <priority>0.75</priority>
    <lastmod>2026-04-02</lastmod>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/the-real-reason-turboquant-shook-the-market-ai-optimization-has-gone-mainstream</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2026-04-02</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/ce5eb80d-1925-42a9-8059-b39cca9a2a75/%E1%84%8B%E1%85%B5%E1%84%8C%E1%85%A2%E1%84%92%E1%85%AE%E1%86%AB1-0+%281%29.jpg</image:loc>
      <image:title>Tech Blog - TurboQuant가 시장을 흔든 진짜 이유: AI 최적화는 이제 메인스트림 - Jaehoon Lee Technical Content Manager, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/b88f09e8-7e0d-4597-a478-3c3262ff0bff/TurboQuant+Content+Figure_01.png</image:loc>
      <image:title>Tech Blog - TurboQuant가 시장을 흔든 진짜 이유: AI 최적화는 이제 메인스트림 - 돋보이게 만드세요</image:title>
      <image:caption>사진1: 중국과 미국의 전력 생산량 비교(2008–2024). 중국은 약 3배 증가한 반면 미국은 정체 상태로, AI 시대 에너지 효율 개선의 필요성을 시사한다. (출처 : AI+HW 2035)</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/03aaa2ac-80d9-4d6f-9692-7aa527d508f2/TurboQuant+Content+Figure_02.png</image:loc>
      <image:title>Tech Blog - TurboQuant가 시장을 흔든 진짜 이유: AI 최적화는 이제 메인스트림 - 돋보이게 만드세요</image:title>
      <image:caption>사진2: NVIDIA HGX B200에서 DeepSeek-R1 서빙 시 정밀도별 처리량 비교. NVFP4(4비트)+MTP 설정이 FP8 대비 약 2배의 처리량을 달성한다. (출처 : NVIDIA Technical Blog)</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/a62033a0-2d70-4cde-a820-068ec29e602a/TurboQuant+Content+Figure_03.png</image:loc>
      <image:title>Tech Blog - TurboQuant가 시장을 흔든 진짜 이유: AI 최적화는 이제 메인스트림 - 돋보이게 만드세요</image:title>
      <image:caption>사진3: KV 캐시 양자화 기법별 LongBench 정확도 비교. TurboQuant는 3.5비트 압축에서도 16비트 Full Cache와 동일한 성능(50.06)을 유지한다. (출처 : Google Research)</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/d193e421-e9a9-41ea-b55d-20be69cd0f91/Design_Techblog_%EC%83%81%EC%84%B8%ED%8E%98%EC%9D%B4%EC%A7%80+%EC%B5%9C%ED%95%98%EB%8B%A8+CTA+%EB%B2%84%ED%8A%BC+%ED%85%9C%ED%94%8C%EB%A6%BF+%EC%A0%9C%EC%9E%91++4%EC%B0%A8%EC%8B%9C%EC%95%88_2.png</image:loc>
      <image:title>Tech Blog - TurboQuant가 시장을 흔든 진짜 이유: AI 최적화는 이제 메인스트림 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/gtc-2026-recap-the-trillion-dollar-inference-race-begins-how-nota-ai-fills-the-gap</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2026-03-23</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/ce5eb80d-1925-42a9-8059-b39cca9a2a75/%E1%84%8B%E1%85%B5%E1%84%8C%E1%85%A2%E1%84%92%E1%85%AE%E1%86%AB1-0+%281%29.jpg</image:loc>
      <image:title>Tech Blog - [GTC 2026 총정리] 1조 달러 추론 경쟁의 시작 : 빈틈을 메우는 노타 - Jaehoon Lee Technical Content Manager, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/777f5bb2-1c1c-48bb-bdc7-921998500134/GTC+2026_01.png</image:loc>
      <image:title>Tech Blog - [GTC 2026 총정리] 1조 달러 추론 경쟁의 시작 : 빈틈을 메우는 노타 - 돋보이게 만드세요</image:title>
      <image:caption>사진1: ChatGPT 등장 이후 2년간 추론 연산 수요가 1만 배 증가한 흐름을 보여주는 슬라이드 (출처: GTC 2026 키노트)</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/4023938e-7007-4821-a79c-57fa0d27ce51/GTC+2026_02.png</image:loc>
      <image:title>Tech Blog - [GTC 2026 총정리] 1조 달러 추론 경쟁의 시작 : 빈틈을 메우는 노타 - 돋보이게 만드세요</image:title>
      <image:caption>사진2: 에이전트의 생성부터 배포, 통제까지 하나로 묶은 NemoClaw 아키텍처 (출처: GTC 2026 키노트)</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/f629ba39-0203-4f64-9f86-7d06f5c34c43/GTC+2026_03.png</image:loc>
      <image:title>Tech Blog - [GTC 2026 총정리] 1조 달러 추론 경쟁의 시작 : 빈틈을 메우는 노타 - 돋보이게 만드세요</image:title>
      <image:caption>사진3: 키노트 무대 위에 모인 로봇, 자율주행 차량, 산업용 중장비 (출처: GTC 2026 키노트)</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/293b1f09-ef10-4d2f-82c3-9f7920eae89c/GTC+2026_04.png</image:loc>
      <image:title>Tech Blog - [GTC 2026 총정리] 1조 달러 추론 경쟁의 시작 : 빈틈을 메우는 노타 - 돋보이게 만드세요</image:title>
      <image:caption>사진4: 프리필을 담당하는 Rubin GPU와 디코드를 담당하는 Groq 3 LPU의 역할 분리 구조(출처: GTC 2026 키노트)</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/genai-everywhere-the-future-of-edge-ai-optimization-with-the-new-netspresso</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2026-03-20</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/a8e2eabf-867b-4c34-acf8-c286b16bb424/image+%285%29.png</image:loc>
      <image:title>Tech Blog - GenAI Everywhere: 새로운 넷츠프레소가 제시하는 Edge AI 최적화의 미래 - NP Product Team, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/0bf1083b-b6a8-4080-b17f-9bd9c22f42af/7C203010.jpg</image:loc>
      <image:title>Tech Blog - GenAI Everywhere: 새로운 넷츠프레소가 제시하는 Edge AI 최적화의 미래 - 돋보이게 만드세요</image:title>
      <image:caption>사진 1. 노타 서울 사무실에서 운영 중인 디바이스팜(Device Farm). 노타에서 성공적으로 AI 모델을 포팅한 기기들로 채워져있다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/db80bb7d-828e-4236-9853-985e2615132c/banner_kor.png</image:loc>
      <image:title>Tech Blog - GenAI Everywhere: 새로운 넷츠프레소가 제시하는 Edge AI 최적화의 미래 - 돋보이게 만드세요</image:title>
      <image:caption>무엇이든, 온라인에서 스토리를 전달하는 방식이 큰 차이를 만들어낼 수 있습니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/52e1fafb-2ba4-477b-a0b1-ca906d7b1cbf/Frame+1597885558_5.png</image:loc>
      <image:title>Tech Blog - GenAI Everywhere: 새로운 넷츠프레소가 제시하는 Edge AI 최적화의 미래 - 돋보이게 만드세요</image:title>
      <image:caption>무엇이든, 온라인에서 스토리를 전달하는 방식이 큰 차이를 만들어낼 수 있습니다.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/notamoequantization-an-moe-specific-quantization-method-for-solar-open-100b-kr</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2026-03-26</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/88b01fdf-fbde-40d9-bf67-18a3c5d93965/%E3%85%87%E3%84%B9%E3%85%87%E3%85%87.png</image:loc>
      <image:title>Tech Blog - NotaMoEQuantization: Solar-Open-100B에 적용된 MoE 구조 특화 양자화 방법론 - Hancheol Park, Ph. D. AI Research Engineer, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/ebdf697a-6345-4402-aa5b-fe3133b4e936/142.+Tairen+Piao.jpg</image:loc>
      <image:title>Tech Blog - NotaMoEQuantization: Solar-Open-100B에 적용된 MoE 구조 특화 양자화 방법론 - Tairen Piao AI Research Engineer, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/5cdc0870-21e2-4ad4-8253-581986d9c24f/37.+%E1%84%80%E1%85%B5%E1%86%B7%E1%84%90%E1%85%A2%E1%84%92%E1%85%A9.jpg</image:loc>
      <image:title>Tech Blog - NotaMoEQuantization: Solar-Open-100B에 적용된 MoE 구조 특화 양자화 방법론 - Tae-Ho Kim CTO &amp; Co-Founder, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/6e826b2b-f385-4898-a8f8-87b84dfa54a6/2.jpg</image:loc>
      <image:title>Tech Blog - NotaMoEQuantization: Solar-Open-100B에 적용된 MoE 구조 특화 양자화 방법론 - ✔️ Resource : K-AI 독자 파운데이션 모델 1차수 통과 모델인 업스테이지 Solar-Open-100B의 공식 양자화 모델: https://huggingface.co/nota-ai/Solar-Open-100B-NotaMoEQuant-Int4</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/2bd74a80-0a01-4db2-953f-1c0860fbc729/%E1%84%8B%E1%85%B5%E1%84%86%E1%85%B5%E1%84%8C%E1%85%B5+%E1%84%8C%E1%85%A1%E1%84%85%E1%85%AD.png</image:loc>
      <image:title>Tech Blog - NotaMoEQuantization: Solar-Open-100B에 적용된 MoE 구조 특화 양자화 방법론 - 돋보이게 만드세요</image:title>
      <image:caption>무엇이든, 온라인에서 스토리를 전달하는 방식이 큰 차이를 만들어낼 수 있습니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/caa843b9-590c-47b6-a4cb-2120e3644e54/%E1%84%89%E1%85%B3%E1%84%8F%E1%85%B3%E1%84%85%E1%85%B5%E1%86%AB%E1%84%89%E1%85%A3%E1%86%BA+2026-03-13+%E1%84%8B%E1%85%A9%E1%84%92%E1%85%AE+8.45.34.png</image:loc>
      <image:title>Tech Blog - NotaMoEQuantization: Solar-Open-100B에 적용된 MoE 구조 특화 양자화 방법론 - 돋보이게 만드세요</image:title>
      <image:caption>무엇이든, 온라인에서 스토리를 전달하는 방식이 큰 차이를 만들어낼 수 있습니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/0033c0e1-8a0f-4d31-87ec-5a12e17ad3bc/%E1%84%89%E1%85%B3%E1%84%8F%E1%85%B3%E1%84%85%E1%85%B5%E1%86%AB%E1%84%89%E1%85%A3%E1%86%BA+2026-03-13+%E1%84%8B%E1%85%A9%E1%84%92%E1%85%AE+8.45.24.png</image:loc>
      <image:title>Tech Blog - NotaMoEQuantization: Solar-Open-100B에 적용된 MoE 구조 특화 양자화 방법론 - 돋보이게 만드세요</image:title>
      <image:caption>무엇이든, 온라인에서 스토리를 전달하는 방식이 큰 차이를 만들어낼 수 있습니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/98c14401-4693-4e17-bf60-7d2e78cef756/%E1%84%89%E1%85%B3%E1%84%8F%E1%85%B3%E1%84%85%E1%85%B5%E1%86%AB%E1%84%89%E1%85%A3%E1%86%BA+2026-03-13+%E1%84%8B%E1%85%A9%E1%84%92%E1%85%AE+8.42.14.png</image:loc>
      <image:title>Tech Blog - NotaMoEQuantization: Solar-Open-100B에 적용된 MoE 구조 특화 양자화 방법론 - 돋보이게 만드세요</image:title>
      <image:caption>무엇이든, 온라인에서 스토리를 전달하는 방식이 큰 차이를 만들어낼 수 있습니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/e720e285-d102-4213-b67c-b4506196c27f/19.png</image:loc>
      <image:title>Tech Blog - NotaMoEQuantization: Solar-Open-100B에 적용된 MoE 구조 특화 양자화 방법론 - 돋보이게 만드세요</image:title>
      <image:caption>무엇이든, 온라인에서 스토리를 전달하는 방식이 큰 차이를 만들어낼 수 있습니다.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/ergo-efficient-high-resolution-visual-understanding-for-vision-language-models</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2026-03-13</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/145e436c-2528-41df-972c-f1b271457df7/Group+40153.jpg</image:loc>
      <image:title>Tech Blog - ERGO: 비전-언어 모델을 위한 효율적인 고해상도 이미지 이해 기술 - Jewon Lee | Wooksu Shin | Seungmin Yang | Ki-Ung Song | Donguk Lim | Jaeyeon Kim | Tae-Ho Kim |  Bo-Kyeong Kim EdgeFM Team, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/515c7af5-c9b5-4024-8d68-c21ad670d529/comparison_with_prior.png</image:loc>
      <image:title>Tech Blog - ERGO: 비전-언어 모델을 위한 효율적인 고해상도 이미지 이해 기술 - 돋보이게 만드세요</image:title>
      <image:caption>무엇이든, 온라인에서 스토리를 전달하는 방식이 큰 차이를 만들어낼 수 있습니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/eba85325-3121-49f2-a14c-699897dc72ad/training_pipeline.png</image:loc>
      <image:title>Tech Blog - ERGO: 비전-언어 모델을 위한 효율적인 고해상도 이미지 이해 기술 - 돋보이게 만드세요</image:title>
      <image:caption>무엇이든, 온라인에서 스토리를 전달하는 방식이 큰 차이를 만들어낼 수 있습니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/14a890f3-400e-45f8-99b2-a41dd32a0ee9/3.webp</image:loc>
      <image:title>Tech Blog - ERGO: 비전-언어 모델을 위한 효율적인 고해상도 이미지 이해 기술 - 돋보이게 만드세요</image:title>
      <image:caption>무엇이든, 온라인에서 스토리를 전달하는 방식이 큰 차이를 만들어낼 수 있습니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/bf9b2678-e654-4aae-bf06-46009229852c/4.webp</image:loc>
      <image:title>Tech Blog - ERGO: 비전-언어 모델을 위한 효율적인 고해상도 이미지 이해 기술 - 돋보이게 만드세요</image:title>
      <image:caption>무엇이든, 온라인에서 스토리를 전달하는 방식이 큰 차이를 만들어낼 수 있습니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/41529e18-7409-4feb-a02b-7813401e6e79/5.webp</image:loc>
      <image:title>Tech Blog - ERGO: 비전-언어 모델을 위한 효율적인 고해상도 이미지 이해 기술 - 돋보이게 만드세요</image:title>
      <image:caption>무엇이든, 온라인에서 스토리를 전달하는 방식이 큰 차이를 만들어낼 수 있습니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/e589d8cf-1b48-4f5a-914d-6a4c771e62be/6.webp</image:loc>
      <image:title>Tech Blog - ERGO: 비전-언어 모델을 위한 효율적인 고해상도 이미지 이해 기술 - 돋보이게 만드세요</image:title>
      <image:caption>무엇이든, 온라인에서 스토리를 전달하는 방식이 큰 차이를 만들어낼 수 있습니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/828cc7ce-a82c-4edc-a569-67039d48579a/7.webp</image:loc>
      <image:title>Tech Blog - ERGO: 비전-언어 모델을 위한 효율적인 고해상도 이미지 이해 기술 - 돋보이게 만드세요</image:title>
      <image:caption>무엇이든, 온라인에서 스토리를 전달하는 방식이 큰 차이를 만들어낼 수 있습니다.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/--ps5jr-rhxmb-kh4r9-pk258</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-12-19</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/9ce2d6c4-079b-47d9-a7c2-c0c4ed388c20/image+%2811%29.png</image:loc>
      <image:title>Tech Blog - NVIDIA Blackwell; NVFP4가 LLM 추론에 미치는 영향 - Seungmin Yang EdgeFM Lead, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/a6eca60e-b52c-4e27-ad66-970d7f71090e/%EC%8A%A4%ED%81%AC%EB%A6%B0%EC%83%B7+2025-10-21+170927.png</image:loc>
      <image:title>Tech Blog - NVIDIA Blackwell; NVFP4가 LLM 추론에 미치는 영향 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/f14ca9e7-1b17-4f1f-9752-19c2bf33da06/%EA%B5%AD%EB%AC%B8+1.png</image:loc>
      <image:title>Tech Blog - NVIDIA Blackwell; NVFP4가 LLM 추론에 미치는 영향 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/e9e40df0-068c-4fb1-9522-53392add535d/%EA%B5%AD%EB%AC%B8+2.png</image:loc>
      <image:title>Tech Blog - NVIDIA Blackwell; NVFP4가 LLM 추론에 미치는 영향 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/b4c2facc-357d-4a2d-84b9-61b6c3d21b61/%EA%B5%AD%EB%AC%B83.png</image:loc>
      <image:title>Tech Blog - NVIDIA Blackwell; NVFP4가 LLM 추론에 미치는 영향 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/717cce77-206b-4c81-bcaa-563cb29a10ee/%EA%B5%AD%EB%AC%B8+4.png</image:loc>
      <image:title>Tech Blog - NVIDIA Blackwell; NVFP4가 LLM 추론에 미치는 영향 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/f16a5bb0-19ab-49e0-a1f9-d916cea7fa60/%EA%B5%AD%EB%AC%B8+5.png</image:loc>
      <image:title>Tech Blog - NVIDIA Blackwell; NVFP4가 LLM 추론에 미치는 영향 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/0e3dc98e-880e-4a6d-8b68-03dcbc8a36a9/TTFT+01.png</image:loc>
      <image:title>Tech Blog - NVIDIA Blackwell; NVFP4가 LLM 추론에 미치는 영향 - Make it stand out</image:title>
      <image:caption>입력 토큰 길이(1024)가 고정된 TTFT-배치 크기 그래프입니다. 왼쪽 그림은 A100에서의 결과를, 오른쪽 그림은 RTX PRO 6000에서의 결과를 나타냅니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/a5aa7491-1ef5-4eba-8740-ab28a9e910c4/TTFT+02.png</image:loc>
      <image:title>Tech Blog - NVIDIA Blackwell; NVFP4가 LLM 추론에 미치는 영향 - Make it stand out</image:title>
      <image:caption>배치 크기(16)가 고정된 TTFT-입력 토큰 길이 그래프입니다. 왼쪽 그림은 A100에서의 결과를, 오른쪽 그림은 RTX PRO 6000에서의 결과를 나타냅니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/591fb43e-86cd-4fc7-935d-7473504f0704/TPOT1.png</image:loc>
      <image:title>Tech Blog - NVIDIA Blackwell; NVFP4가 LLM 추론에 미치는 영향 - Make it stand out</image:title>
      <image:caption>입력 토큰 길이(1024)가 고정된 TPOT-배치 크기 그래프입니다. 왼쪽 그림은 A100에서의 결과를, 오른쪽 그림은 RTX PRO 6000에서의 결과를 나타냅니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/8a9304b8-d92c-4b58-9fc1-a9c4346fdebf/TPOT2.png</image:loc>
      <image:title>Tech Blog - NVIDIA Blackwell; NVFP4가 LLM 추론에 미치는 영향 - Make it stand out</image:title>
      <image:caption>배치 크기(16)가 고정된 처리량-입력 토큰 길이 그래프입니다. 왼쪽 그림은 A100에서의 결과를, 오른쪽 그림은 RTX PRO 6000에서의 결과를 나타냅니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/0697402a-c3d1-4f41-b56c-d0c01d76f721/%EA%B7%B8%EB%A6%BC3.png</image:loc>
      <image:title>Tech Blog - NVIDIA Blackwell; NVFP4가 LLM 추론에 미치는 영향 - Make it stand out</image:title>
      <image:caption>입력 토큰 길이(1024)가 고정된 처리량-배치 크기 그래프입니다. 실선은 RTX PRO 6000 에서의 최적 설정, 점선은 A100 에서의 최적 설정을 의미합니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/b3d06723-7ac2-4700-9994-91da2578f59e/%EA%B7%B8%EB%A6%BC4.png</image:loc>
      <image:title>Tech Blog - NVIDIA Blackwell; NVFP4가 LLM 추론에 미치는 영향 - Make it stand out</image:title>
      <image:caption>배치 크기(16)가 고정된 처리량-입력 토큰 길이 그래프입니다. 실선은 RTX PRO 6000 에서의 최적 설정, 점선은 A100 에서의 최적 설정을 의미합니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/93f2189e-f7b2-4163-85f5-1b844bfb58c4/benchmark_winogrande.png</image:loc>
      <image:title>Tech Blog - NVIDIA Blackwell; NVFP4가 LLM 추론에 미치는 영향 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/a9c23ee4-ac26-485b-8499-a5e202a6e7b2/benchmark_piqa.png</image:loc>
      <image:title>Tech Blog - NVIDIA Blackwell; NVFP4가 LLM 추론에 미치는 영향 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/27450e1e-dec3-465f-87a6-80c73ae8a62e/benchmark_mmlu_5_shot.png</image:loc>
      <image:title>Tech Blog - NVIDIA Blackwell; NVFP4가 LLM 추론에 미치는 영향 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/ce87861a-7d65-4323-8575-463214a62794/benchmark_mmlu_0_shot.png</image:loc>
      <image:title>Tech Blog - NVIDIA Blackwell; NVFP4가 LLM 추론에 미치는 영향 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/66c7afb7-ac77-416d-b296-8ba28a3b7ab0/benchmark_hellaswag.png</image:loc>
      <image:title>Tech Blog - NVIDIA Blackwell; NVFP4가 LLM 추론에 미치는 영향 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/--ps5jr-rhxmb</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-10-21</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/395e24d2-e420-4a7b-a4aa-62a1f44963e8/Marcel.png</image:loc>
      <image:title>Tech Blog - 지식 증류(Self-distillation)를 활용한 비디오 자기지도학습(Self-supervised Learning) 프레임워크 - Marcel Simon, Ph. D. ML Researcher, Nota AI GmbH</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/18cd0c3d-b655-4aa7-bb61-79d15cbb51e2/cto.jpg</image:loc>
      <image:title>Tech Blog - 지식 증류(Self-distillation)를 활용한 비디오 자기지도학습(Self-supervised Learning) 프레임워크 - Tae-Ho Kim CTO &amp; Co-Founder, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/d27ecc58-9c72-4959-a9a5-770bdba499ad/Seul-ki.jpg</image:loc>
      <image:title>Tech Blog - 지식 증류(Self-distillation)를 활용한 비디오 자기지도학습(Self-supervised Learning) 프레임워크 - Seul-Ki Yeom, Ph. D. Research Lead, Nota AI GmbH</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/69e6bf57-8620-40f5-be8f-e75600830219/1.png</image:loc>
      <image:title>Tech Blog - 지식 증류(Self-distillation)를 활용한 비디오 자기지도학습(Self-supervised Learning) 프레임워크 - Make it stand out</image:title>
      <image:caption>그림 1. 학습 개요</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/b1d51ee1-3420-464a-b7e5-665056adae95/2.png</image:loc>
      <image:title>Tech Blog - 지식 증류(Self-distillation)를 활용한 비디오 자기지도학습(Self-supervised Learning) 프레임워크 - Make it stand out</image:title>
      <image:caption>표 1. ADE20K에서의 의미론적 분할 결과</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/6b7b0833-b26b-4619-912c-c7b6025e6506/3.png</image:loc>
      <image:title>Tech Blog - 지식 증류(Self-distillation)를 활용한 비디오 자기지도학습(Self-supervised Learning) 프레임워크 - Make it stand out</image:title>
      <image:caption>표 2. COCO에서의 객체 검출 성능 비교</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/2e73fe12-ac86-4a8d-b6f5-ff84595a690d/4.png</image:loc>
      <image:title>Tech Blog - 지식 증류(Self-distillation)를 활용한 비디오 자기지도학습(Self-supervised Learning) 프레임워크 - Make it stand out</image:title>
      <image:caption>그림 2. 예측 스트라이드 ∆에 따른 ADE20K fast-linear 정확도 변화</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/--ps5jr-n3r37-4dsk9</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-07-10</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/a9af755b-c74d-44c6-add7-3b174be21112/Group+40158.png</image:loc>
      <image:title>Tech Blog - SplitQuant: 엣지 AI 기기에서의 신경망 저비트 양자화를 위한 레이어 분할 기법 - Jaewoo Song Software Engineer, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/4dd235d9-ed69-4035-8ca5-0720a85a41ad/%EC%9E%AC%EC%9A%B0%EB%8B%98+%ED%85%8C%ED%81%AC%EB%B8%94%EB%A1%9C%EA%B7%B8+%EC%82%BD%EB%8F%84.jpeg</image:loc>
      <image:title>Tech Blog - SplitQuant: 엣지 AI 기기에서의 신경망 저비트 양자화를 위한 레이어 분할 기법 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/9aa64427-0d5f-4049-b213-2c1fb1cd2768/1.jpeg</image:loc>
      <image:title>Tech Blog - SplitQuant: 엣지 AI 기기에서의 신경망 저비트 양자화를 위한 레이어 분할 기법 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/--ps5jr-nmlr4</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-05-16</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/145e436c-2528-41df-972c-f1b271457df7/Group+40153.jpg</image:loc>
      <image:title>Tech Blog - LLaMA-3.2-Vision 경량화를 위한 크로스 어텐션 기반의 시각 정보 축소 - Jewon Lee | Ki-Ung Song | Seungmin Yang | Donguk Lim | Jaeyeon Kim | Wooksu Shin | Bo-Kyeong Kim | Tae-Ho Kim EdgeFM Team, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/fb79f560-4613-4374-ab04-588e22bd9d51/yongjaelee-Nov2019.jpg</image:loc>
      <image:title>Tech Blog - LLaMA-3.2-Vision 경량화를 위한 크로스 어텐션 기반의 시각 정보 축소 - Yong Jae Lee, Ph. D. Associate Professor, UW-Madison</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/cfb4a0df-c778-44b3-99b9-805d5731eb2f/2.png</image:loc>
      <image:title>Tech Blog - LLaMA-3.2-Vision 경량화를 위한 크로스 어텐션 기반의 시각 정보 축소 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/67ddf3a8-6cf3-47ee-8ef2-b4bea1a42264/3.png</image:loc>
      <image:title>Tech Blog - LLaMA-3.2-Vision 경량화를 위한 크로스 어텐션 기반의 시각 정보 축소 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/b2b6ea2d-5f49-45b6-af82-958a056ecfac/5.png</image:loc>
      <image:title>Tech Blog - LLaMA-3.2-Vision 경량화를 위한 크로스 어텐션 기반의 시각 정보 축소 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/--ps5jr-n3r37</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-12-18</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/e9c811c2-00fe-4044-a0c6-c2569fa80a5e/%E1%84%89%E1%85%B3%E1%86%AF%E1%84%80%E1%85%B5_%E1%84%89%E1%85%A1%E1%84%8C%E1%85%B5%E1%86%AB.jpeg</image:loc>
      <image:title>Tech Blog - UniForm: 자원이 제한된 엣지 디바이스에서 효율적인 트랜스포머를 위한 재사용 어텐션 메커니즘 - Seul-Ki Yeom, Ph. D. Research Lead, Nota AI GmbH</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/5c288bfa-46a7-4a1f-98d9-8bfe85df630d/CTO_Profile.jpeg</image:loc>
      <image:title>Tech Blog - UniForm: 자원이 제한된 엣지 디바이스에서 효율적인 트랜스포머를 위한 재사용 어텐션 메커니즘 - Tae-Ho Kim CTO &amp; Co-Founder, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/120c64fd-c720-4825-8956-75b1ce59939f/Screenshot+2025-04-08+at+11.10.46+AM.png</image:loc>
      <image:title>Tech Blog - UniForm: 자원이 제한된 엣지 디바이스에서 효율적인 트랜스포머를 위한 재사용 어텐션 메커니즘 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/ea27e513-5bd3-45c3-ac84-165f16ddb134/Screenshot+2025-04-08+at+11.13.07+AM.png</image:loc>
      <image:title>Tech Blog - UniForm: 자원이 제한된 엣지 디바이스에서 효율적인 트랜스포머를 위한 재사용 어텐션 메커니즘 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/--ps5jr</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-05-07</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167987-MVR2JGKPIWEM1YH7YGKM/Performance_%EB%B0%95%ED%95%9C%EC%B2%A0.jpg</image:loc>
      <image:title>Tech Blog - 다국어 대형 언어 모델 생성 텍스트 탐지 연구 - Hancheol Park, Ph. D. AI Research Engineer, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167984-9K4AJFGIVZHAKMTDIYN3/Netspresso_%EA%B9%80%EA%B1%B4%EB%AF%BC_%EC%BB%AC%EB%9F%AC.jpg</image:loc>
      <image:title>Tech Blog - 다국어 대형 언어 모델 생성 텍스트 탐지 연구 - Geonmin Kim, Ph. D. AI Research Engineer, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/892b44e2-e5a7-41dc-88f5-7e596f86b83a/jaeyeonkim.jpeg</image:loc>
      <image:title>Tech Blog - 다국어 대형 언어 모델 생성 텍스트 탐지 연구 - Jaeyeon Kim AI Research Engineer, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/2c17d83d-373d-4e77-95af-4c1de2fbbb50/graph.png</image:loc>
      <image:title>Tech Blog - 다국어 대형 언어 모델 생성 텍스트 탐지 연구 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/d05ece84-aba8-499f-ba41-01046fb013f6/presentation.jpg</image:loc>
      <image:title>Tech Blog - 다국어 대형 언어 모델 생성 텍스트 탐지 연구 - Make it stand out</image:title>
      <image:caption>그림 1. COLING 2025 DAIGenC 워크숍의 Shared Task 1 순위 발표에서 노타가 수상자로 발표되는 장면.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/855a8c4a-f5cf-48f8-85f5-e0a60f86b179/numbers.png</image:loc>
      <image:title>Tech Blog - 다국어 대형 언어 모델 생성 텍스트 탐지 연구 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/4092da7e-da81-44d3-86ff-984b42417e62/graph.png</image:loc>
      <image:title>Tech Blog - 다국어 대형 언어 모델 생성 텍스트 탐지 연구 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/465fc5e7-6cba-4078-9e66-92feb8eef536/result.png</image:loc>
      <image:title>Tech Blog - 다국어 대형 언어 모델 생성 텍스트 탐지 연구 - Make it stand out</image:title>
      <image:caption>표 2. Shared Task 1의 상위 7개 팀 순위표</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/-</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-05-07</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167987-MVR2JGKPIWEM1YH7YGKM/Performance_%EB%B0%95%ED%95%9C%EC%B2%A0.jpg</image:loc>
      <image:title>Tech Blog - 대형 언어 모델은 데이터의 애매성을 어디에서 식별할까? - Hancheol Park, Ph. D. AI Research Engineer, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167984-9K4AJFGIVZHAKMTDIYN3/Netspresso_%EA%B9%80%EA%B1%B4%EB%AF%BC_%EC%BB%AC%EB%9F%AC.jpg</image:loc>
      <image:title>Tech Blog - 대형 언어 모델은 데이터의 애매성을 어디에서 식별할까? - Geonmin Kim, Ph. D. AI Research Engineer, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/c814711c-2901-4348-95f1-9fe321cf39b5/%EC%88%98%EC%A0%95%EA%B0%80%EB%8A%A5%ED%95%9C_%EC%9D%B4%EB%AF%B8%EC%A7%80.png</image:loc>
      <image:title>Tech Blog - 대형 언어 모델은 데이터의 애매성을 어디에서 식별할까? - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/eed2fbe0-4df2-4005-bbf7-bd0dfce1d2d5/image+%2824%29.png</image:loc>
      <image:title>Tech Blog - 대형 언어 모델은 데이터의 애매성을 어디에서 식별할까? - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/4ab952d0-b158-4981-93bd-9046818de908/image+%2825%29.png</image:loc>
      <image:title>Tech Blog - 대형 언어 모델은 데이터의 애매성을 어디에서 식별할까? - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/35400806-f3fa-46f2-9a41-6c13a33d3432/image+%2826%29.png</image:loc>
      <image:title>Tech Blog - 대형 언어 모델은 데이터의 애매성을 어디에서 식별할까? - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/3d269267-cd9f-4b6b-8c30-e0eadd67180a/image+%2827%29.png</image:loc>
      <image:title>Tech Blog - 대형 언어 모델은 데이터의 애매성을 어디에서 식별할까? - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/50cd8736-4a1b-4f47-8b0b-1b85747f9957/image+%2828%29.png</image:loc>
      <image:title>Tech Blog - 대형 언어 모델은 데이터의 애매성을 어디에서 식별할까? - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/deploying-an-efficient-vision-language-model-on-mobile-devices-2krwr</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-05-07</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167664-1DUODJRSSD6YYXPX3JUY/NOTA_Netspresso_%EA%B9%80%EC%9E%AC%EC%97%B0_%EC%BB%AC%EB%9F%AC.jpg</image:loc>
      <image:title>Tech Blog - 모바일 디바이스에 효율적인 비전 언어 모델 배포하기 - Jaeyeon Kim Research Engineer, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167667-RXYOOW7BRYJDN2E8GF9E/Netspresso_%EA%B9%80%EA%B1%B4%EB%AF%BC_%EC%BB%AC%EB%9F%AC.jpg</image:loc>
      <image:title>Tech Blog - 모바일 디바이스에 효율적인 비전 언어 모델 배포하기 - Geonmin Kim Research Engineer, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167670-R76A2AWJQ7XUCK18MM0K/Performance_%EB%B0%95%ED%95%9C%EC%B2%A0.jpg</image:loc>
      <image:title>Tech Blog - 모바일 디바이스에 효율적인 비전 언어 모델 배포하기 - Hancheol Park Team Lead of NetsPresso Application, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167673-4E5127EOV9IBE4DLBK7C/1+%EC%88%98%EC%A0%95.png</image:loc>
      <image:title>Tech Blog - 모바일 디바이스에 효율적인 비전 언어 모델 배포하기 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167676-EUV5N2SX621UMQVFRJ3V/2.png</image:loc>
      <image:title>Tech Blog - 모바일 디바이스에 효율적인 비전 언어 모델 배포하기 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167679-KLYEIG65WD8I091D8U7K/3.png</image:loc>
      <image:title>Tech Blog - 모바일 디바이스에 효율적인 비전 언어 모델 배포하기 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167682-85QY51DQ0942TCQJVKYX/240723_table1_img.png</image:loc>
      <image:title>Tech Blog - 모바일 디바이스에 효율적인 비전 언어 모델 배포하기 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/cluster-self-refinement-for-enhanced-online-multi-camera-people-tracking-2zfks</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-05-07</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167691-KG6X6V50FBPDLNX7AWHW/Performance_%EA%B9%80%EC%A0%95%ED%98%B8.jpg</image:loc>
      <image:title>Tech Blog - 향상된 온라인 다중 카메라 인물 추적을 위한 Cluster Self-Refinement - Jeongho Kim Research Engineer, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167694-XFJ8J9VT58JNBJAFADYY/1.png</image:loc>
      <image:title>Tech Blog - 향상된 온라인 다중 카메라 인물 추적을 위한 Cluster Self-Refinement - Make it stand out</image:title>
      <image:caption>그림 1. 여러 카메라에 걸쳐 사람들을 동일한 아이디로 매핑하여 추적하는 MCPT의 예시를 보여줍니다. 중앙에 있는 이미지는 각 카메라에서 캡처한 사람들의 위치를 2D 맵으로 보여주며, 숫자는 이들의 글로벌 아이디를 나타냅니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167697-TIGA0170Y77LEOFDZ684/2.png</image:loc>
      <image:title>Tech Blog - 향상된 온라인 다중 카메라 인물 추적을 위한 Cluster Self-Refinement - Make it stand out</image:title>
      <image:caption>그림 2. 시스템의 전체 아키텍처입니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167700-OC0ET6TFX0GDWO09VPE7/3.png</image:loc>
      <image:title>Tech Blog - 향상된 온라인 다중 카메라 인물 추적을 위한 Cluster Self-Refinement - Make it stand out</image:title>
      <image:caption>그림 3. CSR의 개요입니다. 왼쪽 그림은 외형 특징을 정제하는 과정을 보여줍니다. 병합 클러스터링을 활용하여 서로 다른 사람들이 동일 클러스터에 저장되었는지 확인한 후, 맞는 경우 클러스터 트랙렛 내의 외형 특징을 정제합니다. 오른쪽 그림은 한 인물이 두 개 이상의 글로벌 ID를 가진 경우를 처리하는 중첩 클러스터 정제 과정을 보여줍니다. CSR 절차는 그림에서 빨간색 원으로 표시된 것처럼 주기적으로 수행됩니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167703-R7U2NPSKM6G8SD0B1X4A/4.png</image:loc>
      <image:title>Tech Blog - 향상된 온라인 다중 카메라 인물 추적을 위한 Cluster Self-Refinement - Make it stand out</image:title>
      <image:caption>표 1. CSR과 EUP를 사용한 소거 연구의 결과입니다. CSR은 Cluster Self-Refinement, EUP는 Enhanced Utilizing Pose estimation을 나타냅니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167706-Q3LLBY6HLUHSXCNOB3B8/5.png</image:loc>
      <image:title>Tech Blog - 향상된 온라인 다중 카메라 인물 추적을 위한 Cluster Self-Refinement - Make it stand out</image:title>
      <image:caption>표 2. Challenge Track 1의 공개 리더보드입니다.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/road-object-detection-robust-to-distorted-objects-at-the-edge-regions-of-images-b4a84</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-05-07</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167714-ZEZGCHFV3BORE8QVICS2/NOTA_Performance_%EC%8B%A0%EC%9A%B1%EC%88%98.jpg</image:loc>
      <image:title>Tech Blog - 이미지 가장자리 왜곡에 특화된 도로 객체 탐지 - Wooksu Shin Research Engineer, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167717-4BFHMEF5X2BAT1YW9E68/1.png</image:loc>
      <image:title>Tech Blog - 이미지 가장자리 왜곡에 특화된 도로 객체 탐지 - Make it stand out</image:title>
      <image:caption>그림 1. 슬라이스 추론의 예시입니다. 위 이미지에서 빨간 상자로 표시된 영역은 모델의 입력 크기에 맞게 크기가 조정된 후 입력됩니다. 그 결과,노란 상자 안의 작은 객체들이 크게 확대되어 모델이 목표 객체를 더 정확하게 탐지할 수 있게 됩니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167720-MGQE8312OV11M199TIGF/2.png</image:loc>
      <image:title>Tech Blog - 이미지 가장자리 왜곡에 특화된 도로 객체 탐지 - Make it stand out</image:title>
      <image:caption>그림 2. (a)는 왜곡된 비목표 객체의 잘못된 예측을 보여줍니다. 도로 표지판이 왜곡되어 차량의 윤곽과 시각적으로 유사하게 보이면서 탐지기가 이를 차량으로 잘못 탐지하는 경우가 있습니다. 그러나 이러한 비목표 객체를 학습한 이후, (b)에서 보이는 것처럼 탐지 오류 문제가 해결되었습니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167723-098325YRDIO812PBJDZS/3.png</image:loc>
      <image:title>Tech Blog - 이미지 가장자리 왜곡에 특화된 도로 객체 탐지 - Make it stand out</image:title>
      <image:caption>그림 3.  히스토그램 평활화를 통해 픽셀 분포가 변화된 모습입니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167726-RJRM5OBW039A7ANJ5RH1/4.png</image:loc>
      <image:title>Tech Blog - 이미지 가장자리 왜곡에 특화된 도로 객체 탐지 - Make it stand out</image:title>
      <image:caption>표 1. 이번 연구에서 사용된 결합된 탐지기들. Swin-L (Liu et al., 2021)과 ViT-L (Dosovitskiy et al., 2021)은 Detection Transformer (DETR) 모델의 기반 구조를 나타내며 Self-Distilled with No Labels (DINO) (Zhang et al., 2023)는 DETR 계열의 아키텍처입니다. Co-DINO (Swin-L)은 Objects365 (Shao et al., 2019)와 Common Objects in Context (COCO) (Lin et al., 2014) 데이터셋으로 사전 훈련되었고 Co-DINO (ViT-L)은 Objects365와 LVIS로 사전 훈련되었습니다. 모든 모델은 FishEye8K (Gochoo et al., 2023) 데이터셋으로 미세 조정되었습니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167730-EC439YN72VIJSQRA67NT/5.png</image:loc>
      <image:title>Tech Blog - 이미지 가장자리 왜곡에 특화된 도로 객체 탐지 - Make it stand out</image:title>
      <image:caption>그림 4. WBF의 예시입니다. 위의 두 이미지는 서로 다른 모델에 의해 예측된 바운딩 박스를 보여줍니다. 아래의 이미지는 WBF를 사용하여 이 바운딩 박스들이 단일 박스로 결합된 모습을 보여줍니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167733-LC2XN233YARK9BUZUVMV/6.png</image:loc>
      <image:title>Tech Blog - 이미지 가장자리 왜곡에 특화된 도로 객체 탐지 - Make it stand out</image:title>
      <image:caption>표 2. 슬라이스 추론과 semi-supervied 학습 방법을 사용한 제거 실험의 결과입니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167735-C91I22WERR1MFHP18E1Z/7.png</image:loc>
      <image:title>Tech Blog - 이미지 가장자리 왜곡에 특화된 도로 객체 탐지 - Make it stand out</image:title>
      <image:caption>표 3. 2024 AI City Challenge Track 4의 공개 Top 10 리더보드입니다.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/edgefusion-on-device-text-to-image-generation-byr62</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-05-07</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167742-LRDR78AREPFSMQW8MTHY/%EB%B0%95%ED%83%9C%EC%9E%84%EB%8B%98.jpg</image:loc>
      <image:title>Tech Blog - EdgeFusion: 온디바이스 텍스트 기반 이미지 생성 - Tairen Piao Research Engineer, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167745-E6RM4W94BU85LQIHAH0M/Edge+Fusion_Figure+1.jpg</image:loc>
      <image:title>Tech Blog - EdgeFusion: 온디바이스 텍스트 기반 이미지 생성 - Make it stand out</image:title>
      <image:caption>그림 1. 텍스트 기반 이미지 생성 결과입니다. 개선된 데이터로 훈련된 EdgeFusion은 몇 단계의 디노이징만으로도 높은 품질의 이미지를 생성할 수 있습니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167748-CIRYC4IAXJ1BW4MLSOQO/Edge+Fusion_Figure+2.jpg</image:loc>
      <image:title>Tech Blog - EdgeFusion: 온디바이스 텍스트 기반 이미지 생성 - Make it stand out</image:title>
      <image:caption>그림 2. 단계가 감소된 컴팩트한 SD. (a) 기본 LCM 적용: BK-LCM-Tiny를 BK-SDM-Tiny의 가중치로 초기화한 후, 샘플링 단계를 줄이기 위해 증류 방식으로 훈련합니다. (b) 우리의 접근법: 더 나은 교사 모델을 통해 LCM의 학생 모델 초기화를 효과적으로 개선합니다. 또한, LCM 훈련 단계에서 원래의 교사 모델을 사용하면 성능이 향상됩니다. 두 단계 모두 고품질 데이터를 활용하는 것이 중요합니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167752-9CIYYL4KOYX98K0RUFHN/Edge+Fusion_Figure+3.jpg</image:loc>
      <image:title>Tech Blog - EdgeFusion: 온디바이스 텍스트 기반 이미지 생성 - Make it stand out</image:title>
      <image:caption>표 1. 다양한 데이터셋에서 훈련된 BK-SDM-Tiny 아키텍처의 결과입니다. COCO 데이터셋에서 25단계로 평가되었습니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167754-OB7NWUPKBHCQB2QIK2PM/Edge+Fusion_Figure+4.jpg</image:loc>
      <image:title>Tech Blog - EdgeFusion: 온디바이스 텍스트 기반 이미지 생성 - Make it stand out</image:title>
      <image:caption>표 2. 인간 선호도 평가 결과입니다. 개선된 데이터와 학생 모델 미세 조정이 없는 동일한 아키텍처와 비교했을 때의 우리 모델의 승률이 보고되었습니다 (1500회 비교, 21명 참가)</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167758-J5SPIK1LILL30W5FAHM6/Edge+Fusion_Figure+5.jpg</image:loc>
      <image:title>Tech Blog - EdgeFusion: 온디바이스 텍스트 기반 이미지 생성 - Make it stand out</image:title>
      <image:caption>그림 3. 32비트 부동소수점(FP32)과 W8A16 양자화된 EdgeFusion 모델(2단계) 간의 성능을 비교한 그림입니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167760-J9VY0I8NOWJ0V2QIGRNK/Edge+Fusion_Figure+6.jpg</image:loc>
      <image:title>Tech Blog - EdgeFusion: 온디바이스 텍스트 기반 이미지 생성 - Make it stand out</image:title>
      <image:caption>표 3. 모델 수준 분할을 사용하면 추론 시간이 단축된다는 것을 확인할 수 있습니다. 크로스 어텐션 블록에서는 모델 수준 분할을 적용하지 않은 작업과 적용한 작업을 비교하여 상대적인 시간 비율을 계산했습니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167764-5IRN62F87C0YV10P0YZM/Edge+Fusion_Figure+7.png</image:loc>
      <image:title>Tech Blog - EdgeFusion: 온디바이스 텍스트 기반 이미지 생성 - Make it stand out</image:title>
      <image:caption>표 4. Exynos 2400에서 모델 수준 분할을 적용한 경우와 적용하지 않은 경우의 벤치마크입니다.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/ld-pruner-efficient-pruning-of-latent-diffusion-models-using-task-agnostic-insights-gwh9x</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-05-07</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167771-TUTKAXG9SDXUN8CT5DDZ/Thibault.png</image:loc>
      <image:title>Tech Blog - LD-Pruner: Task-Agnostic 인사이트를 활용한 Latent Diffusion Models (LDMs)의 효율적인 가지치기 - Thibault Castells Research Engineer, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167774-8S2MF7I88LQVOA96KVDC/Screenshot+2024-04-11+at+3.07.29%E2%80%AFPM.png</image:loc>
      <image:title>Tech Blog - LD-Pruner: Task-Agnostic 인사이트를 활용한 Latent Diffusion Models (LDMs)의 효율적인 가지치기 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167778-TA6ETH5YPZTT24P1SHUQ/Screenshot+2024-04-11+at+3.45.38%E2%80%AFPM.png</image:loc>
      <image:title>Tech Blog - LD-Pruner: Task-Agnostic 인사이트를 활용한 Latent Diffusion Models (LDMs)의 효율적인 가지치기 - Make it stand out</image:title>
      <image:caption>표 1: MS-COCO 256 X 256 검증 세트에서 다양한 모델들이 생성한 T2I를 비교했습니다. 속도 향상 값은 SD-v1.4를 기준으로 상대적으로 측정되었습니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167781-N081UVV8LILKTHR4BPVS/Screenshot+2024-04-11+at+3.44.30%E2%80%AFPM.png</image:loc>
      <image:title>Tech Blog - LD-Pruner: Task-Agnostic 인사이트를 활용한 Latent Diffusion Models (LDMs)의 효율적인 가지치기 - Make it stand out</image:title>
      <image:caption>Figure 2: Qualitative comparison on zero-shot MS-COCO benchmark on T2I. The results of previous studies were obtained with their official released models.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167784-9986GI8D09WJ6SZGT9HS/Screenshot+2024-04-11+at+3.49.52%E2%80%AFPM.png</image:loc>
      <image:title>Tech Blog - LD-Pruner: Task-Agnostic 인사이트를 활용한 Latent Diffusion Models (LDMs)의 효율적인 가지치기 - Make it stand out</image:title>
      <image:caption>그림 3: CelebA-HQ 256 X 256 데이터셋에서 UIG 작업을 위한 훈련 과정 중 프레셰 인셉션 거리의 변화와 두 가지 다른 압축 비율에 대한 결과를 나타냅니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167787-JTN8CT9TPWP7473P47AT/Screenshot+2024-04-11+at+3.54.54%E2%80%AFPM.png</image:loc>
      <image:title>Tech Blog - LD-Pruner: Task-Agnostic 인사이트를 활용한 Latent Diffusion Models (LDMs)의 효율적인 가지치기 - Make it stand out</image:title>
      <image:caption>표 2: AudioDiffusion을 사용한 UAG 작업에서의 압축 성능을 보여줍니다. 12k 단계 동안 미세 조정을 진행했습니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167790-BFBD23I6A6TV5T8KMYQY/Screenshot+2024-04-11+at+4.01.54%E2%80%AFPM.png</image:loc>
      <image:title>Tech Blog - LD-Pruner: Task-Agnostic 인사이트를 활용한 Latent Diffusion Models (LDMs)의 효율적인 가지치기 - Make it stand out</image:title>
      <image:caption>표 3: CelebA-HQ 256 X 256에서 UIG 작업을 위한  압축된 모델(31개의 연산자 수정)이 처음부터 훈련된 경우와 가중치 보존 후 훈련된 경우의 프레셰 인셉션 거리 점수를 나타냅니다. 두 경우 모두 동일한 훈련이 적용되었습니다. 원래 모델의 프레셰 인셉션 거리는 13.85입니다.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/shortened-llm-a-simple-depth-pruning-for-large-language-models-3xfkr</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-05-07</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167797-3EEM9FC65AALZ204YJ66/%EB%B3%B4%EA%B2%BD%EB%8B%98.png</image:loc>
      <image:title>Tech Blog - 축약된 LLM: 대형 언어 모델을 위한 간단한 깊이 가지치기 방법 - Bo-Kyeong Kim Senior Researcher, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167800-NLP0QU9O8IWC7WWS5JTT/image+%284%29.png</image:loc>
      <image:title>Tech Blog - 축약된 LLM: 대형 언어 모델을 위한 간단한 깊이 가지치기 방법 - Make it stand out</image:title>
      <image:caption>그림 1. (a) 가지치기 단위를 비교한 것입니다. 너비 가지치기는 데이터 투영과 변환을 위한 가중치 행렬의 크기를 줄입니다. 깊이 가지치기는 트랜스포머 블록 또는 개별 다중 헤드 어텐션 및 대형 언어 모델 모듈을 제거합니다. (b) 엔비디아 H100 GPU에서 가지치기된 LLaMA-7B 모델의 효율성을 보여줍니다. FLAP과 LLM-Pruner의 너비 가지치기에 비해 우리의 깊이 가지치기는 WikiText-2에서 경쟁력 있는 PPL과 함께 더 빠른 추론을 달성하고 지연-처리량 트레이드오프에서 더 나은 결과를 제공합니다 (오른쪽; M: 배치 크기).</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167803-ADQ9OKYD77E727KGK0Y2/image+%285%29.png</image:loc>
      <image:title>Tech Blog - 축약된 LLM: 대형 언어 모델을 위한 간단한 깊이 가지치기 방법 - Make it stand out</image:title>
      <image:caption>그림 2. 저희의 깊이 가지치기 접근 방식입니다. 간단한 지표를 사용해 중요하지 않은 블록을 식별한 후, 일회성 가지치기를 수행하고 간단한 재훈련을 진행합니다. 오른쪽에 있는 LoRA 그림은 Hu et al. [2022]에서 인용한 것입니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167806-XJHWWJBEYEWJQ2TABYI2/pdf_fig_ppl_bookcorpus_v1.png</image:loc>
      <image:title>Tech Blog - 축약된 LLM: 대형 언어 모델을 위한 간단한 깊이 가지치기 방법 - Make it stand out</image:title>
      <image:caption>그림 3. 보정 세트에서 각 트랜스포머 블록의 중도를 추정한 결과입니다. PPL 점수가 낮은 블록은 가지치기 대상으로 선택됩니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167809-NJ2BWCUUICKJXR4YZRXG/image+%286%29.png</image:loc>
      <image:title>Tech Blog - 축약된 LLM: 대형 언어 모델을 위한 간단한 깊이 가지치기 방법 - Make it stand out</image:title>
      <image:caption>그림 4. 가지치기된 LLaMA-1-7B [Touvron et al., 2023]와 Vicuna-v1.3-13B [Chiang et al., 2023]의 결과입니다. Wanda-sp [Sun et al., 2024; An et al., 2024], FLAP [An et al., 2024], LLM-Pruner [Ma et al., 2023]의 너비 가지치기 방법은 추론 효율성을 저하시키는 경향이 있습니다. 반면 우리의 깊이 가지치기 접근 방식은 생성 속도를 향상시키고 제로샷 작업 성능에서도 우수한 경쟁력을 보입니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167812-2VUZJZUPXOSWTHNW0CM7/image+%287%29.png</image:loc>
      <image:title>Tech Blog - 축약된 LLM: 대형 언어 모델을 위한 간단한 깊이 가지치기 방법 - Make it stand out</image:title>
      <image:caption>그림 5. 생성 예시입니다. 'AI가 몇 초 만에 로고를 만들 수 있다'는 문장을 입력하면 가지치기된 모델은 원래 모델과 유사한 결과를 생성합니다.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/integrating-launchx-with-nvidia-tao-toolkit-for-running-on-various-edge-devices-dedx9</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-05-07</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167819-OA8GDGXGUJI089BEMI6S/%E1%84%83%E1%85%A2%E1%84%8C%E1%85%B5+1%402x.jpg</image:loc>
      <image:title>Tech Blog - 엔비디아 TAO Toolkit 결과 모델을 다양한 엣지 장치에서 실행하기 위한 LaunchX 연동 방법 - Hoin Na CoS Tech Part Manager, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167823-ZB5CF9H5NX1GJ6DIN6EH/Screenshot_2023-11-14_at_4.55.44.png</image:loc>
      <image:title>Tech Blog - 엔비디아 TAO Toolkit 결과 모델을 다양한 엣지 장치에서 실행하기 위한 LaunchX 연동 방법 - Make it stand out</image:title>
      <image:caption>그림 1. 엔비디아 TAO Toolkit 구성도입니다. (TAO Toolkit | 엔비디아 Developer)</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167826-FRTWVCEC3KJJW5886BR7/Screenshot_2023-11-14_at_7.26.14.png</image:loc>
      <image:title>Tech Blog - 엔비디아 TAO Toolkit 결과 모델을 다양한 엣지 장치에서 실행하기 위한 LaunchX 연동 방법 - Make it stand out</image:title>
      <image:caption>그림 2. LaunchX 입니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167829-D4242UNFJW6ZRTVTI186/Screenshot_2023-11-14_at_8.21.20.png</image:loc>
      <image:title>Tech Blog - 엔비디아 TAO Toolkit 결과 모델을 다양한 엣지 장치에서 실행하기 위한 LaunchX 연동 방법 - Make it stand out</image:title>
      <image:caption>그림 3. LaunchX를 사용한 MobilenetV2(TF1) 모델의 다양한 장치에서의 벤치마크 결과를 비교한 것입니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167833-RUCZ97IDJCKUMGQA70UE/%EC%8A%A4%ED%81%AC%EB%A6%B0%EC%83%B7+2023-11-14+165817.png</image:loc>
      <image:title>Tech Blog - 엔비디아 TAO Toolkit 결과 모델을 다양한 엣지 장치에서 실행하기 위한 LaunchX 연동 방법 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/how-netspresso-turbocharges-semantic-segmentation-models-srdxw</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-05-07</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167841-TCD3ZI6LGHRDNPPZHIDD/YoonJae+Yang.gif</image:loc>
      <image:title>Tech Blog - 모바일 AI의 혁신 : 실시간 성능 향상을 위해 넷츠프레소로 의미론적 분할 모델을 가속하는 방법 - YoonJae Yang AI Application Developer, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167847-YOFOA7NC0T883T8RWKN0/Netspresso_%E1%84%8B%E1%85%B5%E1%84%92%E1%85%A7%E1%86%BC%E1%84%8C%E1%85%AE%E1%86%AB_%E1%84%8F%E1%85%A5%E1%86%AF%E1%84%85%E1%85%A5.jpg</image:loc>
      <image:title>Tech Blog - 모바일 AI의 혁신 : 실시간 성능 향상을 위해 넷츠프레소로 의미론적 분할 모델을 가속하는 방법 - Hyungjun Lee Research Engineer, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167855-7FB7UH2IAZUL5W0ZKMB4/Article-4_table1.png</image:loc>
      <image:title>Tech Blog - 모바일 AI의 혁신 : 실시간 성능 향상을 위해 넷츠프레소로 의미론적 분할 모델을 가속하는 방법 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167865-3CNVH7TQLL4HSG26VI7I/Article-4_table2.png</image:loc>
      <image:title>Tech Blog - 모바일 AI의 혁신 : 실시간 성능 향상을 위해 넷츠프레소로 의미론적 분할 모델을 가속하는 방법 - Make it stand out</image:title>
      <image:caption>이 결과는 넷츠프레소의 구조적 가지치기를 통해 PIDNet 모델을 압축함으로써 지연 시간이 크게 개선되었음을 보여줍니다. 이러한 최적화는 PIDNet 모델이 모바일 장치에서 화상 회의나 영상 통화와 같은 실시간 응용 프로그램에 더 적합한 선택이 될 수 있도록 도와줍니다. 모바일 최적화를 위한 우리의 여정을 통해, 아무리 계산이 복잡한 AI 모델이더라도 적절한 도구와 기술을 사용하면 모바일 환경에서 효율적으로 실행될 수 있다는 것을 확인했습니다. 이는 실시간 온디바이스 AI 처리를 요구하는 응용 프로그램에 흥미로운 가능성을 열어줍니다.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/empowering-pedestrian-safety-the-obstacle-detection-app-and-ai-model-optimization-73c3h</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-05-07</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167873-5YPHAL5TH7M19IT921EH/YoonJae+Yang.gif</image:loc>
      <image:title>Tech Blog - 장애물 감지 앱과 AI 모델 최적화로 보행자 안전 강화하기 - Yoonjae Yang AI Application Developer, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167876-IARL5UISD0KJYZLL03OC/%EB%8C%80%EC%A7%80+1%403x.png</image:loc>
      <image:title>Tech Blog - 장애물 감지 앱과 AI 모델 최적화로 보행자 안전 강화하기 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/revolutionizing-laundry-symbol-detection-with-ai-model-optimization-streamlining-the-process-for-precise-results-42yf3</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-05-07</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167884-ND6YXMKF1N6EYQ7T7112/%EB%AC%B4%EC%A0%9C-1.gif</image:loc>
      <image:title>Tech Blog - AI 모델 최적화를 활용한 혁신적인 세탁 기호 탐지 솔루션: 정확도를 위한 프로세스의 간소화 - Yoonjae Yang AI Application Developer, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167887-ZNP1ZU3HFO5CXTXGXKT4/Template_Community_Usecasetable_e5d686cc-54ef-4025-b6fb-e0471967f968.png</image:loc>
      <image:title>Tech Blog - AI 모델 최적화를 활용한 혁신적인 세탁 기호 탐지 솔루션: 정확도를 위한 프로세스의 간소화 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/developing-a-pothole-detection-application-for-safe-driving-ysh63</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-05-07</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167895-ZGTG4J83O6X4SVARAY0G/%EB%8C%80%EC%A7%80+1%403x.jpg</image:loc>
      <image:title>Tech Blog - AI 기반 포트홀 감지를 통한 도로 안전 개선: AI 비전 개발자 및 엔지니어를 위한 성능 최적화 - YoonJae Yang AI Application Developer, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167898-179OB0JGP4AMKA68AD0P/Netspresso_%EC%9D%B4%ED%98%95%EC%A4%80_%EC%BB%AC%EB%9F%AC.jpg</image:loc>
      <image:title>Tech Blog - AI 기반 포트홀 감지를 통한 도로 안전 개선: AI 비전 개발자 및 엔지니어를 위한 성능 최적화 - Hyungjun Lee Research Engineer, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167901-HY8K4HBDHDQGKLW3SK4V/Pothole_table+1.png</image:loc>
      <image:title>Tech Blog - AI 기반 포트홀 감지를 통한 도로 안전 개선: AI 비전 개발자 및 엔지니어를 위한 성능 최적화 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167904-TX3529PEWOCCH7JJPC0N/Pothole_table+2.png</image:loc>
      <image:title>Tech Blog - AI 기반 포트홀 감지를 통한 도로 안전 개선: AI 비전 개발자 및 엔지니어를 위한 성능 최적화 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/community/enhancing-real-time-processing-of-yolov5-l-using-pruning-techniques-in-pynetspresso-hczk4</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-12-19</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167919-045DV71CQ7ZXPJBEWZ5A/Netspresso_%EC%9D%B4%ED%98%95%EC%A4%80_%EC%BB%AC%EB%9F%AC.jpg</image:loc>
      <image:title>Tech Blog - 파이넷츠프레소의 가지치기 기법을 활용한YOLOv5-L의 실시간 처리 속도 향상 - Hyungjun Lee Research Engineer, Nota AI</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167922-8N0WOH1ZIW4UEZ9KFN43/Community_1+%282%29.png</image:loc>
      <image:title>Tech Blog - 파이넷츠프레소의 가지치기 기법을 활용한YOLOv5-L의 실시간 처리 속도 향상 - Make it stand out</image:title>
      <image:caption>그림 1. 파이넷츠프레소를 사용하여 최소한의 성능 저하로 실시간 FPS를 달성하는 모습입니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167926-CPQMP7PP0QDJ6NBD4YL0/pr-latency.png</image:loc>
      <image:title>Tech Blog - 파이넷츠프레소의 가지치기 기법을 활용한YOLOv5-L의 실시간 처리 속도 향상 - Make it stand out</image:title>
      <image:caption>그림 2-1. 가지치기 비율과 지연 시간을 나타내는 그래프입니다. (지연 시간은 Jetson Xavier에서 측정됨)</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167929-2E0CGENXTBWMTMD3GE8I/pr-map.png</image:loc>
      <image:title>Tech Blog - 파이넷츠프레소의 가지치기 기법을 활용한YOLOv5-L의 실시간 처리 속도 향상 - Make it stand out</image:title>
      <image:caption>그림 2-2. 가지치기 비율과 mAP를 나타내는 그래프입니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167932-47D93LWJ92EIZNWUDATE/latency-map.png</image:loc>
      <image:title>Tech Blog - 파이넷츠프레소의 가지치기 기법을 활용한YOLOv5-L의 실시간 처리 속도 향상 - Make it stand out</image:title>
      <image:caption>그림 2-3. 지연 시간과 mAP를 나타내는 그래프입니다. (지연 시간은 Jetson Xavier에서 측정됨)</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167935-N0WQYGYJRG36TTYBWTNQ/Summary+table.png</image:loc>
      <image:title>Tech Blog - 파이넷츠프레소의 가지치기 기법을 활용한YOLOv5-L의 실시간 처리 속도 향상 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167938-30RVVH73DM14FMWDIMYH/Table.png</image:loc>
      <image:title>Tech Blog - 파이넷츠프레소의 가지치기 기법을 활용한YOLOv5-L의 실시간 처리 속도 향상 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167941-S9QE5SY0WQUD5B3H785N/Community+article_5.png</image:loc>
      <image:title>Tech Blog - 파이넷츠프레소의 가지치기 기법을 활용한YOLOv5-L의 실시간 처리 속도 향상 - Make it stand out</image:title>
      <image:caption>그림 3. 전통적 가지치기 기법으로는 매개변수를 완전히 제거할 수 없어, 가속화를 달성하기 어렵습니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167945-PJSIWZZX96CF14ATCGJC/Figure+4.png</image:loc>
      <image:title>Tech Blog - 파이넷츠프레소의 가지치기 기법을 활용한YOLOv5-L의 실시간 처리 속도 향상 - Make it stand out</image:title>
      <image:caption>그림 4. DAP는 각 계층 간의 입력/출력 관계를 고려해 매개변수를 완전히 제거함으로써 경량화된 인공 신경망을 구현합니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167948-JF87WOEKMA2DK2PYGTDV/Figure+5+%282%29.png</image:loc>
      <image:title>Tech Blog - 파이넷츠프레소의 가지치기 기법을 활용한YOLOv5-L의 실시간 처리 속도 향상 - Make it stand out</image:title>
      <image:caption>그림 5. 스킵 연결에서 DAP 작업은 의존성을 추적하고 제거함으로써 인공 신경망의 실질적인 가속화를 가능하게 합니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167950-A1II5391QSRDWT0H6BG2/Python_1.png</image:loc>
      <image:title>Tech Blog - 파이넷츠프레소의 가지치기 기법을 활용한YOLOv5-L의 실시간 처리 속도 향상 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167953-WEO0TC4VY6S21UP3MNMB/Python_2.png</image:loc>
      <image:title>Tech Blog - 파이넷츠프레소의 가지치기 기법을 활용한YOLOv5-L의 실시간 처리 속도 향상 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167957-TZFNERMNGJBHV0IE5MZI/Python_3.png</image:loc>
      <image:title>Tech Blog - 파이넷츠프레소의 가지치기 기법을 활용한YOLOv5-L의 실시간 처리 속도 향상 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167959-TO4V0YJ6L42KCOYPQ1K5/Python_6.png</image:loc>
      <image:title>Tech Blog - 파이넷츠프레소의 가지치기 기법을 활용한YOLOv5-L의 실시간 처리 속도 향상 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167963-OFDDJLKDW4VNQH3CEH5P/Figure+6.png</image:loc>
      <image:title>Tech Blog - 파이넷츠프레소의 가지치기 기법을 활용한YOLOv5-L의 실시간 처리 속도 향상 - Make it stand out</image:title>
      <image:caption>그림 6. 결과가 'output_path'에 저장된 모습입니다.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167966-O7JWK3KZHB4QK71EA173/Python_5.png</image:loc>
      <image:title>Tech Blog - 파이넷츠프레소의 가지치기 기법을 활용한YOLOv5-L의 실시간 처리 속도 향상 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167969-KM4O2AY70ZJU9YT7O3AE/Python+7.png</image:loc>
      <image:title>Tech Blog - 파이넷츠프레소의 가지치기 기법을 활용한YOLOv5-L의 실시간 처리 속도 향상 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304167973-LJBBN9K7205Y0P15GLUS/pynp_main.png</image:loc>
      <image:title>Tech Blog - 파이넷츠프레소의 가지치기 기법을 활용한YOLOv5-L의 실시간 처리 속도 향상 - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/notice</loc>
    <changefreq>daily</changefreq>
    <priority>0.75</priority>
    <lastmod>2026-03-30</lastmod>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/notice/260323</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2026-03-25</lastmod>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/notice/260316</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2026-03-25</lastmod>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/notice/260210</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2026-03-19</lastmod>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/notice/20250714</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2026-03-25</lastmod>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/notice/20250407</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2026-03-30</lastmod>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/notice/20250204</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-02-04</lastmod>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/notice/20241213</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2026-03-25</lastmod>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/notice/2024-04-11</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2026-03-25</lastmod>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/notice/2024-03-18</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2026-03-25</lastmod>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/notice/20240102</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2026-03-25</lastmod>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/news</loc>
    <changefreq>daily</changefreq>
    <priority>0.75</priority>
    <lastmod>2025-12-29</lastmod>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/news/krpressrelease2512</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-12-29</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/35662165-99db-4c67-a552-5b1b9b1e241f/Nota+AI_PressRelease_Cover.png</image:loc>
      <image:title>newsroom - 노타, 삼성전자 차세대 AP ‘엑시노스 2600’ AI 최적화 기술 공급… 온디바이스 AI 필수 기술로 자리매김한다! - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/news/pressrelease2511</loc>
    <changefreq>monthly</changefreq>
    <priority>0.5</priority>
    <lastmod>2025-12-28</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/40cb2b15-c316-4308-8ae7-3a022bde213c/Nota+AI_PressRelease_Cover.png</image:loc>
      <image:title>newsroom - 노타, 삼성전자와 ‘엑시노스 AI 최적화 기술 공급’ 계약 “온디바이스 생성형 AI 대중화 이끈다” - Make it stand out</image:title>
      <image:caption>Whatever it is, the way you tell your story online can make all the difference.</image:caption>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/privacy-policy</loc>
    <changefreq>daily</changefreq>
    <priority>0.75</priority>
    <lastmod>2024-12-04</lastmod>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/contact-us</loc>
    <changefreq>daily</changefreq>
    <priority>0.75</priority>
    <lastmod>2025-07-10</lastmod>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/aboutus</loc>
    <changefreq>daily</changefreq>
    <priority>0.75</priority>
    <lastmod>2025-07-11</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219228-M7U8DODKSNSJVWCV7P32/company-background-blue.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219233-6DE5L86PCUSSFID7VQN3/%E1%84%82%E1%85%A9%E1%84%90%E1%85%A1_%E1%84%92%E1%85%AC%E1%84%8B%E1%85%B4%E1%84%89%E1%85%B5%E1%86%AF.JPG</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219236-990H26RICE476R2AMXO3/%E1%84%82%E1%85%A9%E1%84%90%E1%85%A1_%E1%84%8B%E1%85%B5%E1%86%AB%E1%84%86%E1%85%AE%E1%86%AF2.JPG</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219239-7FWKFIVOPZ0KRV6S8P7S/%E1%84%82%E1%85%A9%E1%84%90%E1%85%A1_%E1%84%8B%E1%85%B5%E1%86%AB%E1%84%86%E1%85%AE%E1%86%AF9.JPG</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219242-8IT64HVRAH2XOO1VV6C2/DSC09728.JPG</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219245-IWL80R9Y4MNB1OJ5FGCU/7C203058+%281%29.JPG</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219248-VN8SWFB0TH8ALJXQF9XX/%E1%84%82%E1%85%A9%E1%84%90%E1%85%A1_%E1%84%85%E1%85%A1%E1%84%8B%E1%85%AE%E1%86%AB%E1%84%8C%E1%85%B5.jpg</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219251-CWYUJDP2XLEF2BSKPNDC/%ED%9A%8C%EC%9D%98%EC%8B%A4_CTO+%ED%9A%8C%EC%9D%98%EC%A4%91_.jpg</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219274-GF8WBTGPSYNJ05TB3353/Frame+75.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219342-AF6XK6BIEYQPR2ORRSEW/Frame+69.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219417-TAMGPF6G4Y1R7UIW1GU2/Frame+77.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219303-XXLH7LV3RF9Z8XZRGQ6N/Frame+74.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219358-FP328QUONISKJFJ7E5ZW/Frame+72.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219403-EUTP7MLI75IVOHLDHBTX/Frame+68.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219280-FI3R3312TWH4FIPCMAOJ/Frame+70.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219330-OK15Z8M8R7XEMTVCZIJ1/Frame+79.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219411-KRFNZEZG8I9DBZ716K2O/Frame+82.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219286-56LXB99NOYLSHNOEDYHD/Frame+78.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219353-XC3B7G7D0S10IMFY6B53/Frame+76.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219423-NUG7NZHZH3SVF5M72UTT/Frame+87.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219298-GT0DWG71SUF1LHQ17PCN/Frame+81.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219364-F6MVANEF2BTHQM0OUSMS/Frame+67.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219386-J3634A42BA0WTAIGRZQK/Frame+84.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219292-PSXS5PX0JFI3IOL2KMXG/Frame+85.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219347-EY0IANRVW7APEQ0QQU95/Frame+65.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219397-JWMJ2TTDETGLHBKTIS46/Frame+88.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219261-I4ZXG0PT43HU1AOF9G3Q/Frame+80.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219325-OZR0ESYQLQFSWK8ZWM72/Frame+86.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219391-2U50ZI3QZ1U6SMYMAM7R/Frame+83.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219267-WHWWEYRW5C55XBII118K/Frame+73.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219336-3SV6J6XLGWLX4UG6QCJU/Frame+71.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304219319-B3ZWBF531RQELPPT9Z4P/qualcomm+logo+rgb+blue.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/634f84f0-14df-4612-bf1d-566374d3ab09/Main_partners_FuriosaAI.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/80b323d5-6a78-47d7-861c-19426656408c/pytorch-logo.png</image:loc>
      <image:title>About us</image:title>
    </image:image>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/ccd6206c-b8e2-485d-8fe2-269a90b863cb/a.png</image:loc>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/careers</loc>
    <changefreq>daily</changefreq>
    <priority>0.75</priority>
    <lastmod>2025-07-15</lastmod>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/newsroom</loc>
    <changefreq>daily</changefreq>
    <priority>0.75</priority>
    <lastmod>2026-03-17</lastmod>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/netspresso</loc>
    <changefreq>daily</changefreq>
    <priority>0.75</priority>
    <lastmod>2026-01-23</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/059c95ee-f270-456e-b09d-f1a6cb8ce860/NetsPresso_Basic_Logo_Navy.png</image:loc>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/ai-solutions-notavisionagent</loc>
    <changefreq>daily</changefreq>
    <priority>0.75</priority>
    <lastmod>2026-01-23</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/045ad042-35fb-4148-9380-a8f6093511cd/%40keyvisual.png</image:loc>
    </image:image>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/ai-solutions-industrialsafety</loc>
    <changefreq>daily</changefreq>
    <priority>0.75</priority>
    <lastmod>2026-03-23</lastmod>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/ai-solutions-surveillance</loc>
    <changefreq>daily</changefreq>
    <priority>0.75</priority>
    <lastmod>2025-10-27</lastmod>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/ai-solutions-its</loc>
    <changefreq>daily</changefreq>
    <priority>0.75</priority>
    <lastmod>2025-10-27</lastmod>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/ai-solutions-dms-fr</loc>
    <changefreq>daily</changefreq>
    <priority>0.75</priority>
    <lastmod>2025-10-27</lastmod>
  </url>
  <url>
    <loc>https://www.kr.nota.ai/main</loc>
    <changefreq>daily</changefreq>
    <priority>1.0</priority>
    <lastmod>2026-03-31</lastmod>
    <image:image>
      <image:loc>https://images.squarespace-cdn.com/content/v1/67501f4ff33c0d4f02c21e5f/1733304205481-KRC5NSAC10VUQOTY93U3/KakaoTalk_Photo_2024-11-28-22-17-02.png</image:loc>
    </image:image>
  </url>
</urlset>

