Dilated Convolution with Learnable Spacings makes visual models more aligned with humans: a Grad-CAM study

Open in new window