amlankar / adascan-public Goto Github PK

View Code? Open in Web Editor NEW

46.0 46.0 15.0 95 KB

Code for AdaScan: Adaptive Scan Pooling (CVPR 2017)

License: GNU General Public License v3.0

Python 98.00% Shell 2.00%

action-recognition cvpr-2017 deep-learning tensorflow

adascan-public's People

Contributors

Stargazers

Watchers

Forkers

hawklucky cylhf1995 luohuifen mandaliadhaval leileestone xiaobai12345 shubhampachori12110095 grvsh r05921061 hanimiao xiaoyu5301 zhiweicui guoruiwang josephgesnouin gcxamy

adascan-public's Issues

unable to find file

Hello! the VGG numpy files link is 404,unable to find file！And initialize flow models link is the caffe model ,but The project uses tensorflow！ thank you !!

About the difference with the paper

Hi:

I find the model somehow separates the vgg for RGB and flow apart, while in the paper it seems that there are two vggs, and the RGB and flow sequence go through them and are concated together in a 4096 * 25 matrix. Does that means that the following MLP and classifier only take the input of either RGB or flow into account?

Thanks!

i have a question as follows about training

There is no pre-trained models?

Excuses me, will the pre-trained models release？

Link to VGG numpy files is lost

Thanks for your wonderful code. I met a problem when I try to train from scratch. The link to VGG numpy files seems to be lost. Could you please check it?

Recommend Projects