(ICASSP) INTEGRATED GRAD-CAM: SENSITIVITY-AWARE VISUAL EXPLANATION OF DEEP CONVOLUTIONAL NETWORKS VIA INTEGRATED GRADIENT-BASED SCORING
(WACV2021) F-CAM: Full Resolution Class Activation Maps via Guided Parametric Upscaling
(CVPR2021) Relevance-CAM: Your Model Already Knows Where to Look
https:
//github.com/mongeoroo/Relevance-CAM
(ICCV2021) LFI-CAM: Learning Feature Importance for Better Visual Explanation
出发点:为了利用ABN的注意力机制,又防止ABS生成unreliable visual explanation,LFI-CAM的注意力分支将特征图视为掩码,并为每个特征图获得特征importance confidence(Similar with Score-CAM)
REF
https://www.jianshu.com/p/c3ac3b11f1e0