The weight configuration of the convolutional layer does not match the number of channels in the input data

[changganqi]

I use PNG photo to train the model , but I met this problem. Anyone can help me? Thanks !