[논문 정리] Learning enhanced features and inferring twice for fine-grained image classification

논문정보 Learning enhanced features and inferring twice for fine-grained image classification X-MOL www.x-mol.net GitHub - boxyao/Forcing-Network Contribute to boxyao/Forcing-Network development by creating an account on GitHub. github.com 논문정리 Abstract FGVC는 비슷한 같은 카테고리에 속하는 데이터들의 하위 카테고리를 구별하는 것을 목표로 한다. 현재 연구는 FGVC에서 차별성있는 특징이 중요하다는 것을 입증했지만, 그 외 특징들 또한 올바른 분류에 기여할 수 있다는 것을 간과하였으며, 추출된 특징들은 미묘한 영..

• format_list_bulleted 논문
• · 2023. 8. 31.
• textsms

explainable swin transformer 관련 자료 정리

understand how visualization of herbs works success visualizaing attention map in vit understand what do each feature maps in swin transformer stand for (and also in vit) Applied Sciences | Free Full-Text | Swin Transformer-Based Object Detection Model Using Explainable Meta-Learning Mining (mdpi.com) Swin Transformer-Based Object Detection Model Using Explainable Meta-Learning Mining In order t..

• format_list_bulleted 관련 자료 정리
• · 2023. 6. 4.
• textsms

[논문 정리] Fine-grained Visual Classification with High-temperature Refinement and Background Suppression

논문정보 Fine-grained Visual Classification with High-temperature Refinement and Background Suppression Fine-grained Visual Classification with High-temperature Refinement and Background Suppression Fine-grained visual classification is a challenging task due to the high similarity between categories and distinct differences among data within one single category. To address the challenges, previous ..

• format_list_bulleted 논문
• · 2023. 5. 26.
• textsms

2023-05-01~2023-05-05

보호되어 있는 글입니다.

• format_list_bulleted 개발사항
• · 2023. 4. 29.

Drawing CAM

CAM을 그리는 코드인데 bakward함수가 없어서 내 문제 해결할 수 있을 것 같음 (test시에 gradient를 계산해야 하는 상황.. torch.no_grad를 써야 gpu메모리 문제없이 학습가능해서) cam def cam(model, cam_topk=1, images=None, labels=None, truncate=False, shift=0.0): if images is not None: _ = model.forward(images) if labels is None: _, labels = torch.topk( model.pred, k=cam_topk, dim=1, largest=True, sorted=True) labels = labels[:, [cam_topk-1]] labels = labe..

• format_list_bulleted 파이썬 Python
• · 2023. 4. 25.
• textsms

Extract bounding box from cam

GitHub - won-bae/rethinkingCAM: Official implementation of Rethinking Class Activation Mapping for Weakly Supervised Object Loca Official implementation of Rethinking Class Activation Mapping for Weakly Supervised Object Localization (ECCV 2020) - GitHub - won-bae/rethinkingCAM: Official implementation of Rethinking Class Ac... github.com github 코드에서 cam으로부터 bounding box를 추출하는 부분을 따왔다. 예상은 했다만 코..

• format_list_bulleted 파이썬 Python
• · 2023. 4. 25.
• textsms