Fig. 10From: Understanding deep learning via backtracking and deconvolutionThe values of the common pixelsBack to article page