ICCV 2021 Supplementary submission for Paper ID 7911.

As part of the supplementary material submission, we include code snippets that implement our approach. Upon publication of this work, we will publicly release the full codebase.

0. iccv2021_supplementary_material.pdf
- Contains a write up that provides additional detail necessary for reproducing our results.

1. attention_fusion.py
- Contains our implementation of the Motion Guided Attention Fusion module.

2. resnet_helper.py
- Please see lines 274-275 for how MGAF is incorporated into a ResNet BottleneckTransform layer seamlessly. 

3. video_model_builder.py
- Please see lines after 149 for our overall architecture definition. 
- From lines 380 and on, the forward pass is defined.

4. temporal_model.py
- Contains the temporal convolutional neural network implementation that we use as our objects model.
