Closed ritikraj7 closed 2 years ago
There should be only 'ifmap_height*ifmap_width' unique addresses in ifmap operand matrix because in 2D convolution same input image is used across different filter channels to extract different features of the same image.
There should be only 'ifmap_height*ifmap_width' unique addresses in ifmap operand matrix because in 2D convolution same input image is used across different filter channels to extract different features of the same image.