I was given the following config file to fix pose fs_vid2vid training. I need to know more about the config file and there is zero documentation.
I need to know what all of the hyperparameters mean and what I will need to change and for a base development what I do not need to change.
hyperparams such asimage_save_iter, snapshot_save_epoch, num_epochs_temporal_step, initial_sequence_length, num_videos_to_test, num_frames_per_video, batch_size, etc.
I need to know that given I have different length videos with different resolutions (wxh of image frames) how is this handled --must I do something on my own for pre-processing.
Also num_frames_per_video (is set initially to 10---that seems really low to capture motion).
ALSO given for pose fs_vid2vid you say to use the preprocessing script from the OLD previous repository
We do not plan to provide documentation for the config file as the functions of the variables are self-explained as one traces the code. We also try to make the names as meaningful as much.
I was given the following config file to fix pose fs_vid2vid training. I need to know more about the config file and there is zero documentation.
I need to know what all of the hyperparameters mean and what I will need to change and for a base development what I do not need to change.
hyperparams such as
image_save_iter, snapshot_save_epoch, num_epochs_temporal_step, initial_sequence_length, num_videos_to_test, num_frames_per_video, batch_size, etc.
I need to know that given I have different length videos with different resolutions (wxh of image frames) how is this handled --must I do something on my own for pre-processing.
Also num_frames_per_video (is set initially to 10---that seems really low to capture motion).
ALSO given for pose fs_vid2vid you say to use the preprocessing script from the OLD previous repository