AI9 [paper review] (VPT) Visual Prompt Tuning Visual Prompt Tuning - ECCV 2022https://arxiv.org/abs/2203.12119 Visual Prompt TuningThe current modus operandi in adapting pre-trained models involves updating all the backbone parameters, ie, full fine-tuning. This paper introduces Visual Prompt Tuning (VPT) as an efficient and effective alternative to full fine-tuning for large-scale Trarxiv.org1. Prompt TuingPrompting- The promising approach.. 2024. 7. 21. [논문 리뷰] (SHIP) Improving Zero-Shot Generalization for CLIP with Synthesized Prompts Improving Zero-Shot Generalization for CLIP with Synthesized Prompts - ICCV 2023이번에는 Co-CoOp의 방식을 차용한 논문에 대해 리뷰하도록 하겠습니다.https://arxiv.org/abs/2307.07397 Improving Zero-Shot Generalization for CLIP with Synthesized PromptsWith the growing interest in pretrained vision-language models like CLIP, recent research has focused on adapting these models to downstream tasks. Despite achieving promising .. 2024. 7. 21. [논문 리뷰] (Co-CoOp) Conditional Prompt Learning for Vision-Language Models Conditional Prompt Learning for Vision-Language Models - CVPR 2022 https://arxiv.org/abs/2203.05557v2 Conditional Prompt Learning for Vision-Language ModelsWith the rise of powerful pre-trained vision-language models like CLIP, it becomes essential to investigate ways to adapt these models to downstream datasets. A recently proposed method named Context Optimization (CoOp) introduces the concept.. 2024. 7. 20. [논문 리뷰] (CoOp) Learning to Prompt for Vision-Language Models - IJCV 2022 Learning to Prompt for Vision-Language Models - IJCV 2022https://arxiv.org/abs/2109.01134 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.o.. 2024. 7. 20. [논문 리뷰](Kor.ver) Hyper-class Augmented and Regularized Deep Learning for Fine-grained Image Classification Hyper-class Augmented and Regularized Deep Learning for Fine-grained Image Classification CVPR.2015.7298880 FGIC는 서로 유사한 카테고리 사이의 미세한 차이를 구별하는 것이 목표유사한 클래스들 간의 작은 차이를 식별해야하나 어려움→ 높은 분류 성능을 내기 어려우며 과적합이 발생하기 쉬움기존 연구의 한계 : 전통적인 방법들은 이러한 미세한 차이를 효과적으로 학습하기 어려움대규모 외부 데이터셋(예: ImageNet)에서 딥 CNN을 사전 학습하고 작은 규모의 대상 데이터에서 미세 조정하여 특정 분류 작업에 맞추는 것딥 CNN의 성공에서 중요한 요소는 대규모 라벨이 붙은 학습 데이터에 접근할 수 있는 것세밀한 이미지 분.. 2024. 7. 13. [paper review](Eng.ver) Hyper-class Augmented and Regularized Deep Learning for Fine-grained Image Classification Hyper-class Augmented and Regularized Deep Learning for Fine-grained Image Classification CVPR.2015.7298880 The challenges of FGIC: - Data Scarcity: expensive to obtain a large number of labeled images → Data Augmentation: Identify hyperclasses and acquire a large number of images labeled with hyper-class from easily accessible search engines for multi-task learning. - Large Intra-Class Variat.. 2024. 7. 13. 이전 1 2 다음
