Load pretrained model YOLO_small.ckpt error about yolo_tensorflow

[hizhangp]

You should load the pretrained model in V1, please pay attention to the parameters in save(restorer) operation.

[zihao-lu]

thanks，i figured out this problem. The YOLO_small.ckpt was broken when I downloaded it. Finally I solved it..... Thanks a lot.

[zihao-lu]

I have another problem why the model is so hard to converge. Can you tell me some reasons? Thanks.
Have a good holiday.

[hizhangp]

Because it is a pretrained model, so it has already converged and cannot achieve a lower loss.

[zihao-lu]

hello, another problem, why you have this layer 'net = slim.fully_connected(net, 512, scope='fc_33')'
a FC_layer with 512 output_size it doesn't apear in the paper