stanleyshly / everybodydancenow Goto Github PK

Hello stanleyshly, I am a newbie to computer vision. Could you give me some advice how to run this program on google colab? I mean, when I run the '!python graph_train.py...' and '!python train_fullts.py...', it reports 'FileNotFoundError: [Errno 2] No such file or directory: '.../data/scratch/caroline/frames/jason_frames'' and '...AssertionError: MY_TRAINING_DATASET/train_label is not a valid directory'. I don't know if it is a problem of the datasets or the confirguration. I just want to run it successfully now. I would be very appreciated if you can reply.

Hi stanleyshly, first of all thank you for this repository.

Could you please upload the pretrained models of global, local and face stage?

When I run global stage training command, the MY_TRAINING_DATASET looks for train_facetexts128 train_img train_label folders. But in the official data it has 5 videos. How to train on multiple videos? Should I just put them into the same folder?

Btw, I think you've trained the model in the colab. Can you produce desired outputs? If so could you please share the checkpoint of the model which would save me from some unnecessary power consumption :)

Hi, do you have the working pose normalization code "graph_posenorm.py" ?
I have tried your script and the one from the original folder, but they don't work.
In the Readme, it's written that you would have replaced it with a new working one.

Thank you!

Thanks for your great implementation of EverybodyDaceNow.

Do you successfully reproduce the reported paper result?

hello @stanleyshly
is the model need to be trained on my data first or not?