ActionDetection-DBG icon indicating copy to clipboard operation
ActionDetection-DBG copied to clipboard

Awesome project! Question about rescale the feature length.

Open ideaRunner opened this issue 5 years ago • 15 comments

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.

ideaRunner avatar Nov 15 '19 06:11 ideaRunner

video_info_19993.json?

pengxiaoxiao avatar Nov 15 '19 07:11 pengxiaoxiao

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

pengxiaoxiao avatar Nov 15 '19 08:11 pengxiaoxiao

@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 avatar Nov 15 '19 09:11 swordlidev

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

swordlidev avatar Nov 15 '19 09:11 swordlidev

yean, thank you!

pengxiaoxiao avatar Nov 15 '19 10:11 pengxiaoxiao

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

pengxiaoxiao avatar Nov 15 '19 10:11 pengxiaoxiao

why?

pengxiaoxiao avatar Nov 15 '19 10:11 pengxiaoxiao

@pengxiaoxiao
check your tensorflow version

swordlidev avatar Nov 15 '19 11:11 swordlidev

Tensorflow-gpu == 1.9.0 Python == 3.6

pengxiaoxiao avatar Nov 15 '19 12:11 pengxiaoxiao

@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 avatar Nov 15 '19 13:11 ideaRunner

@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.

ideaRunner avatar Nov 15 '19 13:11 ideaRunner

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

pengxiaoxiao avatar Nov 16 '19 03:11 pengxiaoxiao

@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?

gtgtgt1117 avatar Nov 28 '19 03:11 gtgtgt1117

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

ideaRunner avatar Nov 28 '19 09:11 ideaRunner

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

quanh1990 avatar Apr 01 '20 14:04 quanh1990