search

frostinassiky / gtad

The official implementation of G-TAD: Sub-Graph Localization for Temporal Action Detection
https://www.deepgcns.org/app/g-tad
Apache License 2.0
216 stars 44 forks source link

About THUMOS14 feature #31

Closed souler256 closed 3 years ago

souler256 commented 4 years ago

For THUMOS-14, the video features are extracted using TSN model pre-trained on Kinetics. Can provide THUMOS-14 features extracted by TSN model pre-trained on Anet1.3(Xiong et. al. submit to ActivityNet Challenge 2016) or provide the code which extract THUMOS-14 features. Thanks a lot.

frostinassiky commented 4 years ago

Hi @souler256 . Thanks for asking.

I did not use the feature extracted by TSN model pre-trained on Anet1.3 in my paper. I also do not have the code to extract that feature.

thematrixduo commented 3 years ago

Could you share how where do you get the features from? Is it the low-level feature from Thumos14 website?