Study With Inha

ICLR 2023 Spotlight (notable-top-25%),(SparK) Designing BERT for Convolutional Networks: Sparse and Hierarchical Masked Modeling논문 링크: https://openreview.net/forum?id=NRxydtWup1SGitHub: https://github.com/keyu-tian/SparK Designing BERT for Convolutional Networks: Sparse and Hierarchical...This paper presents a simple yet powerful framework to pre-train convolutional network (convnet) with Sparse..

ICLR 2024 Oral Paper,Is ImageNet worth 1 video?Learning strong image encoders from 1 long unlabelled video논문 링크: https://openreview.net/forum?id=Yen1lGns2o Is ImageNet worth 1 video? Learning strong image encoders from 1...Self-supervised learning has unlocked the potential of scaling up pretraining to billions of images, since annotation is unnecessary. But are we making the best use of data? H..

ICLR 2024, Interpreting CLIP's Image Representation via Text-Based Decomposition 논문 링크: https://openreview.net/attachment?id=5Ca9sSzuDp&name=pdf프로젝트 페이지 링크: https://yossigandelsman.github.io/clip_decomposition/ 1. Introduction최근 많은 논문들에서 거대한 Text, Image Pair로 학습시킨 CLIP 모델을 활용하는 후속 연구들을 진행하고 있다.본 논문에서는 CLIP의 이미지 인코더를 분석하여 모델의 각 구성 요소가 final representation에 미치는 영향을 해석 가능한 텍스트로 설명하고 있다.먼저, Attentio..