Senyao Han

Results 3 comments of Senyao Han

预测的时候模型test annotation

不需要给test annotation, feed_dict中删掉即可

pretrained model with a test image: Invalid Argument and Incompatible Shapes

I believe it's a padding problem. just go to slim/nets/vgg.py, change this line: net = slim.conv2d(net, 4096, [7, 7], padding='VALID', scope='fc6') to padding='SAME'. It will be okay. The thing is,...

got the error Shapes (3, 3, 1, 64) and (64, 1, 5, 5) are incompatible

change the first two kernel size from (3,3) to (5,5) ![image](https://user-images.githubusercontent.com/25421473/65846065-1e419400-e36f-11e9-8e5c-ae2a25888932.png)