ForeSight: Multi-View Streaming Joint Object Detection and Trajectory Forecasting

TL;DR

ForeSight is a novel multi-task framework for autonomous vehicle perception that jointly detects objects and forecasts their trajectories using a streaming, bidirectional learning approach, improving accuracy and temporal consistency without explicit object tracking.

Contribution

It introduces a joint detection and forecasting model with streaming, bidirectional learning, and forecast-aware transformers, eliminating the need for explicit object tracking and enhancing multi-view perception.

Findings

Achieves state-of-the-art EPA of 54.9% on nuScenes

Surpasses previous methods by 9.3% in EPA

Attains best mAP and minADE among multi-view models

Abstract

We introduce ForeSight, a novel joint detection and forecasting framework for vision-based 3D perception in autonomous vehicles. Traditional approaches treat detection and forecasting as separate sequential tasks, limiting their ability to leverage temporal cues. ForeSight addresses this limitation with a multi-task streaming and bidirectional learning approach, allowing detection and forecasting to share query memory and propagate information seamlessly. The forecast-aware detection transformer enhances spatial reasoning by integrating trajectory predictions from a multiple hypothesis forecast memory queue, while the streaming forecast transformer improves temporal consistency using past forecasts and refined detections. Unlike tracking-based methods, ForeSight eliminates the need for explicit object association, reducing error propagation with a tracking-free model that efficiently scales across multi-frame sequences. Experiments on the nuScenes dataset show that ForeSight achieves state-of-the-art performance, achieving an EPA of 54.9%, surpassing previous methods by 9.3%, while also attaining the best mAP and minADE among multi-view detection and forecasting models.