Justin Johnson

Check out my neural-style speed benchmarks: https://github.com/jcjohnson/neural-style#speed The Maxwell Titan X benchmarks used CUDA 7.0 and cuDNN 4 and the Pascal Titan X benchmarks used CUDA 8.0 RC and cuDNN...

I'm guessing your trying to install Torch with a standard Lua install? Usually you install Torch by following the instructions here: http://torch.ch/docs/getting-started.html#_ which ships with its own version of luajit,...

We don't have code for that in this repo, but it's relatively simple. First use the extractFeatures method to get boxes and features for the database images: https://github.com/jcjohnson/densecap/blob/master/densecap/DenseCapModel.lua#L285 Next run...

I have not seen that error before - does it happen consistently or was it a one-time error? Were there other jobs running on the same GPU at the same...

There is some problem with your protobuf which is needed by loadcaffe; I'm not exactly sure how to solve this.

Thanks - those are all indeed small bugs. Do you want to send a pull request to fix them?

Your initial learning rate seems too high - that might be causing it to blow up. Training already takes days on a GPU so I didn't expect anyone to want...

I used 1e-6 for initial learning rate. It's hard to say exactly how much slower CPU will be than GPU, but it could be as much as 10x slower.

The .json checkpoint files saved during training contain training loss history as well as validation set mAP. You can use those to watch the training process.

@htoyryla `U2` is just a nearest-neighbor upsampling; you may get better results if you stick a stride-1 convolution between the two upsampling layers: `-arch c9s1-16,d32,d64,R64,R64,R64,R64,R64,U2,c3s1-32,U2,c9s1-3`