Champion | Lite-QCNet Team wins the 1st in the CVPR 2024 Argoverse 2 Multi-Agent Motion Forecasting Challenge

The Lite-QCNet Team, supervised by Prof. Jianping Wang, wins first place in the CVPR 2024 Argoverse 2 Multi-Agent Motion Forecasting Challenge.Multi-agent Motion Forecasting using the Argoverse 2 Motion Forecasting Dataset. Given the position, orientation, and category of actors in a scene, predict the future motion of several key actors in the future.

The motion forecasting model utilized in this challenge was developed by Prof. Jianping Wang and her PhD student, Mr. Zikang Zhou, in collaboration with researchers from the Hon Hai Research Institute. This innovative model enhances computing and memory efficiency by incorporating a patch-based self-attention mechanism into the sequence modeling process. Remarkably, it reduces latency by approximately 45% compared to the team’s winning solution in the 2023 Argoverse 2 challenge, without sacrificing prediction accuracy. This advancement provides new insights into deploying large models within constrained computing resources.