Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Number of epoch #4

Open
sudonto opened this issue Nov 14, 2018 · 2 comments
Open

Number of epoch #4

sudonto opened this issue Nov 14, 2018 · 2 comments

Comments

@sudonto
Copy link

sudonto commented Nov 14, 2018

Hi, I am wondering how many epoch you set for training optical flow images on UCF101? And how many days the training process was? I read your code in flowUCF.py, you nb of epoch is 50, but I never had a good result.

Anyway, thanks for the code

@sudonto
Copy link
Author

sudonto commented Nov 14, 2018

And also, this repo doesnt have file named trainlistfull.pkl as written in flowUCF.py. Could you share this file? And also the weight: keras-ucf101-TVL1flow-resnet50-split1.hdf5 since I cannot find this in your google drive. Sorry for asking too much :)

EDIT: forget about the weight, I know it now :)

@pfabreu
Copy link
Owner

pfabreu commented Nov 14, 2018

Thanks for your interest and im glad you found the initialization wegihts, I may have to change some names. I've been a bit busy but I will add the .pkl soon. The number of epochs is set in each file (I think as nb_epoch). The training on UCF was done for the whole network without frozing any layers, try that. On UCF101-24 the training is quite fast, around a few hours but on the AVA dataset it takes two days, but I still want to test if any alternatives exist.

I'll keep you updated on the next few days.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants