silent_videos

This release contains silent top-view videos of each test sequence. You can overlay these videos with binaural audio generated by your model to generate top-view visualizations similar to those used in our supplemental video.

pretrained_models_v1.0

This release contains two pre-trained binaural networks:

1. a small model with a single WaveNet block for faster experiments;
2. a large model with three WaveNet blocks as in the ICLR paper.

binaural_dataset_v1.0

Download the binaural dataset here.