search

EasonXiao-888 / UVCOM

[CVPR 2024] Bridging the Gap: A Unified Video Comprehension Framework for Moment Retrieval and Highlight Detection
MIT License
73 stars 4 forks source link

How can i inference with a single video and get its output after downloading the checkpoints? #7

Open whycantfindaname opened 3 months ago

whycantfindaname commented 3 months ago

Splendid work! But I wonder that how can i inference with a single video and get its output after downloading the checkpoints?

whycantfindaname commented 3 months ago

And I find there are two sklearn package in requirements.txt. When I used pip install -r requirements.txt it reported error as The 'sklearn' PyPI package is deprecated, use 'scikit-learn' rather than 'sklearn' for pip commands. The bug is in the 73 line and 80 line of the requirements.txt

scikit-learn==1.0.2
scipy==1.7.3
seaborn==0.12.2
sentry-sdk==1.12.1
setproctitle==1.3.2
shortuuid==1.0.11
six==1.16.0
sklearn==0.0.post5
EasonXiao-888 commented 3 months ago

@whycantfindaname Thanks for your interest! The video running demo could refer to run on video. What's more, thanks for your notice, I will update the requirements.txt and delete the sklearn==0.0.post5! ❤️

whycantfindaname commented 3 months ago

@whycantfindaname Thanks for your interest! The video running demo could refer to run on video. What's more, thanks for your notice, I will update the requirements.txt and delete the sklearn==0.0.post5! ❤️

I have seen the https://github.com/wjun0830/QD-DETR/tree/main/run_on_video and I wonder what should I change to make it run for this project UVCOM because I notice that some functions there are not in your code? And how to change the checkpoint path to use your checkpoint? Looking foward to your reply!

awkrail commented 3 months ago

@whycantfindaname @EasonXiao-888 Hi, this is not official from the authors, but I implemented an inference APIs for moment retrieval and highlight detection. It implement UVCOM, and enables you to run single video inference. If you are interested, please try our library. https://github.com/line/lighthouse

whycantfindaname commented 3 months ago

@whycantfindaname @EasonXiao-888 Hi, this is not official from the authors, but I implemented an inference APIs for moment retrieval and highlight detection. It implement UVCOM, and enables you to run single video inference. If you are interested, please try our library. https://github.com/line/lighthouse

Thank you! I will try your library as soon as possible and provide feedback！