Open mrgloom opened 7 years ago
Why upsampling right after pooling? Why not to add one more convolution between?
https://github.com/alexgkendall/SegNet-Tutorial/blob/master/Models/segnet_basic_train.prototxt#L260
Also why Relu activations not used in upsampling blocks?
Why upsampling right after pooling? Why not to add one more convolution between?
https://github.com/alexgkendall/SegNet-Tutorial/blob/master/Models/segnet_basic_train.prototxt#L260