Wei Yang
Wei Yang
Hmm, seems this problem cannot be reproduced. Would you mind to train again and see whether everything goes well?
I think 4 images are not sufficient to train a pose estimator.
Interesting. What is the batchsize you used during training and testing? It is possible that the BN is not trained properly.
Thanks. Would you be able to create a PR?
Hi @Chinese-xingyi ,Did you follow these commands for training your model? https://github.com/bearpaw/pytorch-classification/blob/master/TRAINING.md#cifar-100
Thanks for the PR. I really appreciate your contribution of upgrading the code to Pytorch 0.4 and 1.0. However, there is part of the contributions that I do not plan...
Hi. Currently, I have no machine to run the experiments. Let's see if any one can help.
Hi Kanza, Can you try the following command to install qlua? `luarocks install qtlua` You need to install all the dependencies before running the code: https://github.com/bearpaw/pose-attention#installation
pose-int means using intermedia supervision. This part of code is borrowed from a history version of [stacked hourglass network](https://github.com/umich-vl/pose-hg-train).
Sorry to bother but this work is AWESOME especially for the online editor!!!!!