CV166 [2026-1] 정유림 - A Simple Framework for Contrastive Learning of Visual Representations Title: A Simple Framework for Contrastive Learning of Visual RepresentationsCitations: 33,428 (as of July 11, 2026)Link: https://arxiv.org/pdf/2002.05709ICML 2020Previous approach to learn visual representationGenerative approachesA generative method attempts to model or reconstruct the input data.SimCLR paper says that pixel-generation can be computationally expensive and may not be necessary f.. 2026. 7. 11. [2026-1] 김지원 - Denoising Diffusion Probabilistic Models 논문 소개 : Denoising Diffusion Probabilistic Models (DDPM)Jonathan Ho, Ajay Jain, Pieter Abbeel (UC Berkeley) · NeurIPS 2020arXiv: 2006.11239 · 공식 구현들어가며요즘 이미지 생성의 표준이 되어버린 디퓨전 모델, 그 출발점이 된 논문임. 사실 디퓨전 모델 자체는 2015년 Sohl-Dickstein et al.이 비평형 열역학(nonequilibrium thermodynamics)에서 아이디어를 빌려와 처음 제안했음. 다만 그때까진 "정의는 깔끔하고 학습도 잘 되는데, 정작 고품질 샘플을 만들어낸 적은 없는" 모델이었음.이 논문의 기여를 한 줄로 요약하면 이렇게 정리됨.디퓨전 모델이 실제로 GAN급,.. 2026. 6. 12. [2026-1] 이루가 - What does CLIP know about a red circle? Visual prompt engineering for VLM 논문 링크: https://arxiv.org/abs/2304.06712 What does CLIP know about a red circle? Visual prompt engineering for VLMsLarge-scale Vision-Language Models, such as CLIP, learn powerful image-text representations that have found numerous applications, from zero-shot classification to text-to-image generation. Despite that, their capabilities for solving novel discriminativearxiv.org ABSTRACT이 논문은 CLIP .. 2026. 4. 25. [2026-1] 정유림 - DataComp: In search of the next generation of multimodal datasets 논문 : https://arxiv.org/abs/2304.14108 DataComp: In search of the next generation of multimodal datasetsMultimodal datasets are a critical component in recent breakthroughs such as Stable Diffusion and GPT-4, yet their design does not receive the same research attention as model architectures or training algorithms. To address this shortcoming in the ML ecosarxiv.org 보통 머신러닝 benchmark는 데이터셋을 고정.. 2026. 4. 25. 이전 1 2 3 4 ··· 42 다음
