DEFT: Detection Embeddings for Tracking

Abstract

Most modern multiple object tracking (MOT) systems follow the tracking-by-detection paradigm, consisting of a detector followed by a method for associating detections into tracks. There is a long history in tracking of combining motion and appearance features to provide robustness to occlusions and other challenges, but typically this comes with the trade-off of a more complex and slower implementation. Recent successes on popular 2D tracking benchmarks indicate that top-scores can be achieved using a state-of-the-art detector and relatively simple associations relying on single-frame spatial offsets -- notably outperforming contemporary methods that leverage learned appearance features to help re-identify lost tracks. In this paper, we propose an efficient joint detection and tracking model named DEFT, or "Detection Embeddings for Tracking." Our approach relies on an appearance-based object matching network jointly-learned with an underlying object detection network. An LSTM is also added to capture motion constraints. DEFT has comparable accuracy and speed to the top methods on 2D online tracking leaderboards while having significant advantages in robustness when applied to more challenging tracking data. DEFT raises the bar on the nuScenes monocular 3D tracking challenge, more than doubling the performance of the previous top method. Code is publicly available.

arXiv commentAccepted at CVPR 2021, ADP3 Workshop on Autonomous Driving: Perception, Prediction and Planning

References in corpus (12)

• Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
• YOLOv3: An Incremental Improvement
• Simple Online and Realtime Tracking
• R-FCN: Object Detection via Region-based Fully Convolutional Networks
• MOT16: A Benchmark for Multi-Object Tracking
• Deep Learning in Video Multi-Object Tracking: A Survey
• A Survey on Multi-Task Learning
• Multi-Task Learning Using Uncertainty to Weigh Losses for Scene Geometry and Semantics
• Class-balanced Grouping and Sampling for Point Cloud 3D Object Detection
• Joint Monocular 3D Vehicle Detection and Tracking
• 3D Multi-Object Tracking: A Baseline and New Evaluation Metrics
• End-to-end Learning Improves Static Object Geo-localization in Monocular Video

Cited by in corpus (7)

• A Comprehensive Review of Computer Vision in Sports: Open Issues, Future Trends and Research Directions
• DeepFusionMOT: A 3D Multi-Object Tracking Framework Based on Camera-LiDAR Fusion with Deep Association
• TripletTrack: 3D Object Tracking using Triplet Embeddings and LSTM
• YOLO11-JDE: Fast and Accurate Multi-Object Tracking with Self-Supervised Re-ID
• Split and Connect: A Universal Tracklet Booster for Multi-Object Tracking
• Multi-object Tracking with Tracked Object Bounding Box Association
• Motion Prediction Performance Analysis for Autonomous Driving Systems and the Effects of Tracking Noise