본문 바로가기

쓰기

 
발표자 김누리 
발표일자 2022-06-29 
저자 Pillai, Vipin, et al. 
학회명 CVPR 2022 
논문지  
Post-hoc explanation methods, e.g., Grad-CAM, enable humans to inspect the spatial regions responsible for a particular network decision. However, it is shown that such explanations are not always consistent with human priors, such as consistency across image transformations. Given an interpretation algorithm, e.g., Grad-CAM, we introduce a novel training method to train the model to produce more consistent explanations. Since obtaining the ground truth for a desired model interpretation is not a well-defined task, we adopt ideas from contrastive self-supervised learning, and apply them to the interpretations of the model rather than its embeddings. We show that our method, Contrastive Grad-CAM Consistency (CGC), results in Grad-CAM interpretation heatmaps that are more consistent with human annotations while still achieving comparable classification accuracy. Moreover, our method acts as a regularizer and improves the accuracy on limited-data, fine-grained classification settings. In addition, because our method does not rely on annotations, it allows for the incorporation of unlabeled data into training, which enables better generalization of the model. The code is available here: https://github.com/UCDvision/CGC

    2022

      Federated Multi-Target Domain Adaptation
      2022.09.07
      발표자: 강용훈     발표일자: 2022-09-07     저자: Chun-Han Yao, Boqing Gong, Hang Qi, Yin Cui, Yukun Zhu, Ming-Hsuan Yang     학회명: WACV 2022    
      How Much More Data Do I Need?
      2022.08.24
      발표자: 지혜림     발표일자: 2022-08-24     저자: Rafid Mahmood1 James Lucas1,2,3 David Acuna1,2,3 Daiqing Li1 Jonah Philion1,2,3 Jose M. Alvarez1 Zhiding Yu1 Sanja Fidler1,2,3 Marc T. Law1 1NVIDIA 2University of Toronto 3Vector Institute     학회명: CVPR 2022