HPNet: Dynamic Trajectory Forecasting with Historical Prediction Attention

1 code implementation • 9 Apr 2024 • Xiaolong Tang, Meina Kan, Shiguang Shan, Zhilong Ji, Jinfeng Bai, Xilin Chen

The proposed Historical Prediction Attention together with the Agent Attention and Mode Attention is further formulated as the Triple Factorized Attention module, serving as the core design of HPNet. Experiments on the Argoverse and INTERACTION datasets show that HPNet achieves state-of-the-art performance, and generates accurate and stable future trajectories.

Autonomous Driving Trajectory Forecasting

Filter Pruning based on Information Capacity and Independence

no code implementations • 7 Mar 2023 • Xiaolong Tang, Tianheng Hu, Yufeng Shi

We introduce information capacity, a metric that represents the amount of information contained in a filter.