search

bryanyzhu / Hidden-Two-Stream

Caffe implementation for "Hidden Two-Stream Convolutional Networks for Action Recognition"
Other
196 stars 68 forks source link

Question about the init temporal model #1

Closed whwu95 closed 6 years ago

whwu95 commented 6 years ago

In this paper, you said "the temporal model is initialized with the snapshot provided by Wang",could you give me the snapshot URL? Thank you.

bryanyzhu commented 6 years ago

Sure, here it is.

https://github.com/yjxiong/caffe/tree/action_recog/models/action_recognition

bryanyzhu commented 6 years ago

oh, i mean the second url.

But i also try the first url as init model. It just take longer time (around 8K iter) to reach the same performance.

cherishtttz commented 6 years ago

hello, when I compiled the code and I followed the 'README' to make under Ubuntu16.04, I can't compile it successfully, the mistake is 'recipe for target '.build_release/tools/convert_imageset_and_disparity.o failed', I don't know how to solve this problem, have you ever met that question? Please help me ,thanks so much!

bryanyzhu commented 6 years ago

Hi, I think the problem is related to this file here. You can follow the error messages to debug. If there is no detailed error message, I think you can also delete this cpp file from the tools directory. It shouldn't effect our functionality. The cpp file is made for original FlowNet and DispNet paper.