TimoSaemann

Could you try it without comment the resize params? new_height: 512 new_width: 512

The _cityscapes_weights.caffemodel_ has changed blob size, because bn layers are merged. For finetuning you need the weights before the bn layer were merged. I have uploaded this weights (cityscapes_weights_before_bn_merge.caffemodel).

Have a look into the tutorial: https://github.com/TimoSaemann/ENet/tree/master/Tutorial

Hi @xiupingmath, since I am just interested in traffic scenes, I have not trained ENet on indoor datasets, but it should straight forward to Cityscapes.

Have a look in the tutorial on how to measure the runtime layer-by-layer.

I do not encounter errors while "make runtest". Using Ubuntu 14.04, Cuda 8.0, cuDNN v4, v5, v5.1, Titan X. How did you compile it? Make or CMake? Cuda version? cuDNN...

I can not reproduce that error. I tried it on 3 different machines and no error occurred: 1. Ubuntu 14.04, CUDA 8.0, Titan X (Pascal), cuDNN v.4 /v.5 /v.5.1, compiled...

https://github.com/TimoSaemann/caffe-segnet-cudnn5/issues/10

Hi Yi, A few thoughts from my side: 1. When the line is commented out, the default bn_mode is active which is "LEARN". In this mode it normalizes the layer...

Yes that's correct. The behaviour you describe I've already noticed when I trained with 'bn_mode: INFERENCE'. However, this does not lead to the best possible quality. I also want to...