Visual and auditory perception is crucial for observing the world. When we hear a sound, our brain will extract semantic information and locate the sounding source.

We also propose a loss function adapted to the severe class imbalance. We show that our model outperforms state-of-the-art alternatives on UA V point clouds.