search

hongluzhou / composer

Compositional Reasoning of Group Activity in Videos with Keypoint-Only Modality
31 stars 5 forks source link

Q about ablations in paper #6

Closed ajtao closed 1 year ago

ajtao commented 1 year ago

Hello, congrats on the amazing work!

For Table 2 of the Supplementary material, it shows that 1-scale keypoint model achieved 91.2 test accuracy. I was wondering if you can say a little bit about what portion of the model this was. Was this a model that only used the first transformer layer "inner" and not the other 3 transformer layers (middle, outer, group)?

hongluzhou commented 1 year ago

Thank you! ☺️ Sorry about my late reply! Yes, this ablation variant only used the first transformer layer "inner". Please see Eq. (6) and Eq. (7) in the supplementary material for the exact formulation:

1 scale

Feel free to let me know if you have any follow-up questions!

Best, Honglu