RTFM icon indicating copy to clipboard operation
RTFM copied to clipboard

Published 20 hours ago • verified publisher icon tianyu0207

Can you share the code on XD-Violence?

Open wjtan99 opened this issue 2 years ago • 5 comments

Hi, thanks for sharing your great work. Can you share you code of RTFM on XD-Violence or the Sultani on XD-Violence? I could not reproduce the results on XD-Violence?

wjtan99 avatar Jun 18 '22 17:06 wjtan99

Never mind, I changed your code and the AP was very close to your result in the paper.

wjtan99 avatar Jun 18 '22 22:06 wjtan99

Hi, I can repeat your RTFM on XD-Violence I3d features with very close AP = 0.764 (not your 0.778 yet). However I cannot repeat Sultani on XD-Violence I3d features. Your paper says 0.7568, I can only get 0.7. Did you have your own code or use Sultani's original code? This code should be very simple, however, I just cannot get better results. Can you share the code if you have your own implementation? Thanks a lot.

wjtan99 avatar Jun 21 '22 22:06 wjtan99

Hi, I can repeat your RTFM on XD-Violence I3d features with very close AP = 0.764 (not your 0.778 yet). However I cannot repeat Sultani on XD-Violence I3d features. Your paper says 0.7568, I can only get 0.7. Did you have your own code or use Sultani's original code? This code should be very simple, however, I just cannot get better results. Can you share the code if you have your own implementation? Thanks a lot.

hello, I'm wondering how do you obtain your I3D features? cause I tried to use the I3D feats extraction rep for the purpose but further test10crops on obtained features gave me very different results from what I obtain with ready feature files. My relevant issue on the topic

chamecall avatar Jul 05 '22 16:07 chamecall

Hi, I got around 0.61 AP for violence dataset. Would you please share about how you modify the original codes? Are you using 5-crop I3D features provided by the dataset author? Or are you using 10-crop I3D features extracted by yourself?

coranholmes avatar Oct 05 '22 09:10 coranholmes

Hi, I can repeat your RTFM on XD-Violence I3d features with very close AP = 0.764 (not your 0.778 yet). However I cannot repeat Sultani on XD-Violence I3d features. Your paper says 0.7568, I can only get 0.7. Did you have your own code or use Sultani's original code? This code should be very simple, however, I just cannot get better results. Can you share the code if you have your own implementation? Thanks a lot.

hello, I'm wondering how do you obtain your I3D features? cause I tried to use the I3D feats extraction rep for the purpose but further test10crops on obtained features gave me very different results from what I obtain with ready feature files. My relevant issue on the topic

Hi would you please share your AP on violence dataset?

coranholmes avatar Oct 05 '22 09:10 coranholmes