search

Tencent / ActionDetection-DBG

Code for AAAI2020 paper "Fast Learning of Temporal Action Proposal via Dense Boundary Generator"
Other
348 stars 103 forks source link

Awesome project! Question about rescale the feature length. #1

Open ideaRunner opened 5 years ago

ideaRunner commented 5 years ago

Thanks for sharing! The inference speed is extremely faster with a better result. I'm thinking about how to reproduce the result on other datasets. Could you share some detail or insight about how to use linear interpolation to rescale the feature length of all videos to same length 100 ?

Any method function, reference papers or codes will be thankful.

pengxiaoxiao commented 5 years ago

video_info_19993.json?

pengxiaoxiao commented 5 years ago

Hello, do you have the code to reproduce this article? The author has no working training code. How do you reason?

swordlidev commented 5 years ago

@ideaRunner After you use TSN to extract video features, the features length is different. You must reshape these features by fitting and interpolation functions. image

swordlidev commented 5 years ago

Hi @pengxiaoxiao With these inference codes，you can repeat the results in our paper.

pengxiaoxiao commented 5 years ago

yean, thank you!

pengxiaoxiao commented 5 years ago

tensorflow.python.framework.errors_impl.NotFoundError: /home/shawxiao/video/ActionDetection-DBG/custom_op/prop_tcfg.so: undefined symbol: _ZN10tensorflow8internal21CheckOpMessageBuilder9NewStringEv

pengxiaoxiao commented 5 years ago

why?

swordlidev commented 5 years ago

@pengxiaoxiao
check your tensorflow version

pengxiaoxiao commented 5 years ago

Tensorflow-gpu == 1.9.0 Python == 3.6

ideaRunner commented 5 years ago

@lijiannuist Yes, even some video's feature-length is less than 100, it must be fixed-length to feed into the network. Thank you for sharing such kind of straightforward code!

ideaRunner commented 5 years ago

@pengxiaoxiao Hi, pengxiaoxiao. For now, I am going to reproduce the inference results. It is better to re-open an issue as your question is not related to this one. The more detailed information about this issue, the more accurate the solution you will get.

pengxiaoxiao commented 5 years ago

thank you! I solved it by g++-5!

gtgtgt1117 commented 5 years ago

@ideaRunner After you use TSN to extract video features, the features length is different. You must reshape these features by fitting and interpolation functions. image

How to extract features with TSN? Is removing the FC layer of TSN a feature?

ideaRunner commented 5 years ago

hi @gtgtgt1117, The author answered at https://github.com/TencentYoutuResearch/ActionDetection-DBG/issues/6

quanh1990 commented 4 years ago

hi @gtgtgt1117, The author answered at https://github.com/TencentYoutuResearch/ActionDetection-DBG/issues/6

@ideaRunner Page missing, could you please describe the solution, thanks very much