My Vision, Computer Vision

Learning to Prompt for Vision-Language ModelsLarge pre-trained vision-language models like CLIP have shown great potential in learning representations that are transferable across a wide range of downstream tasks. Different from the traditional representation learning that is based mostly on discretiarxiv.org발행일 : 2021년 9월 2일저널/학회 : SPRINGER 2022ProblemCLIP과 같은 기존 VLM은 프롬프트 엔지니어링을 통한 Zero-shot t..

LLaMA: Open and Efficient Foundation Language ModelsWe introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, withoarxiv.org발행일 : 2023. 02. 27.Meta AIProblem최근, 한정된 컴퓨터 예산(Budget)에서 LLM과 데이터셋 크기에 대한 최적화 연구..

BEiT: BERT Pre-Training of Image TransformersWe introduce a self-supervised vision representation model BEiT, which stands for Bidirectional Encoder representation from Image Transformers. Following BERT developed in the natural language processing area, we propose a masked image modeling task to prearxiv.orgProblemBEIT(Bidirectional Encoder representation from Image Transformers)를 제안한다.ViT는 CNN..

BERT: Pre-training of Deep Bidirectional Transformers for Language UnderstandingWe introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models, BERT is designed to pre-train deep bidirectional representations from unlaarxiv.org발행일 : 2018. 10. 11.Google AI LanguageProblem기존의 Langu..

BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language ModelsThe cost of vision-and-language pre-training has become increasingly prohibitive due to end-to-end training of large-scale models. This paper proposes BLIP-2, a generic and efficient pre-training strategy that bootstraps vision-language pre-training from oarxiv.org Problem대부분의 State-of-the-art V..

CLIPPO: Image-and-Language Understanding from Pixels OnlyMultimodal models are becoming increasingly effective, in part due to unified components, such as the Transformer architecture. However, multimodal models still often consist of many task- and modality-specific pieces and training procedures. For example,arxiv.orgProblem대부분의 멀티모달 모델은 모달리티 별로 구성 요소가 다르고, 다른 태스크에 적용하기 위해 그에 맞는 추가적인 학습 절차가 필요..