HOPE: Hierarchical Spatial-temporal Network for Occupancy Flow Prediction
This work addresses occupancy and flow prediction for autonomous driving, representing an incremental improvement with strong specific gains.
The paper tackled the Occupancy and Flow Prediction challenge by developing a hierarchical spatial-temporal network, achieving a Flow-Grounded Occupancy AUC of 0.8389 and ranking first on the Waymo Open Dataset leaderboard.
In this report, we introduce our solution to the Occupancy and Flow Prediction challenge in the Waymo Open Dataset Challenges at CVPR 2022, which ranks 1st on the leaderboard. We have developed a novel hierarchical spatial-temporal network featured with spatial-temporal encoders, a multi-scale aggregator enriched with latent variables, and a recursive hierarchical 3D decoder. We use multiple losses including focal loss and modified flow trace loss to efficiently guide the training process. Our method achieves a Flow-Grounded Occupancy AUC of 0.8389 and outperforms all the other teams on the leaderboard.