# foundationvision/bytetrack **Attribution required: if you use, quote, or summarise this content, you must credit and link back to [awesome-repositories.com](https://awesome-repositories.com/repository/foundationvision-bytetrack).** 6,492 stars · 1,123 forks · Python · MIT ## Links - GitHub: https://github.com/FoundationVision/ByteTrack - awesome-repositories: https://awesome-repositories.com/repository/foundationvision-bytetrack.md ## Topics `deployment` `multi-object-tracking` `pytorch` `real-time` ## Description ByteTrack is a multi-object tracking framework that implements the ByteTrack algorithm, an ECCV 2022 method designed to recover occluded objects and reduce trajectory fragmentation. The core innovation of the project is its association algorithm, which processes every detection box—including low-confidence ones—by using separate high and low score thresholds, Kalman filter motion prediction, and Hungarian algorithm matching to produce consistent object identities across video frames. The project distinguishes itself by its comprehensive approach to handling occlusions and fragmented trajectories through low-score detection recovery, where low-confidence detections are re-matched with unmatched tracklets. It includes a complete tracklet lifecycle management system that creates, updates, and deletes tracklets based on association results and a configurable maximum lost frame count. ByteTrack also provides model export capabilities, allowing trained tracking models to be converted to ONNX, TensorRT, ncnn, or DeepStream formats for production inference deployment. The framework supports custom tracker training through a fine-tuning pipeline for adapting pretrained detectors to custom multi-object tracking datasets. It processes video files frame by frame, integrating detection results from any object detector to produce bounding boxes and tracked object identities as output. The project includes ONNX model export and TensorRT inference optimization with FP16 quantization for low-latency GPU inference. ## Tags ### Artificial Intelligence & ML - [Video Object Tracking](https://awesome-repositories.com/f/artificial-intelligence-ml/video-object-tracking.md) — Tracks multiple objects across video frames by associating all detection boxes to maintain consistent identities. - [Detector-Agnostic Tracking Integrations](https://awesome-repositories.com/f/artificial-intelligence-ml/computer-vision-systems/computer-vision/object-detection-tracking/object-detection/detector-agnostic-tracking-integrations.md) — Integrates detection results from any object detector into the tracking algorithm for consistent object identities. - [Detection-to-Track Association](https://awesome-repositories.com/f/artificial-intelligence-ml/computer-vision-systems/computer-vision/object-detection-tracking/object-tracking-systems/detection-to-track-association.md) — Implements the ByteTrack association algorithm that matches every detection box to existing track IDs. - [ByteTrack Association Algorithms](https://awesome-repositories.com/f/artificial-intelligence-ml/computer-vision-systems/computer-vision/object-detection-tracking/object-tracking-systems/detection-to-track-association/bytetrack-association-algorithms.md) — Implements the ByteTrack ECCV 2022 algorithm that recovers occluded objects by associating all detection boxes. - [Detector-Agnostic Trackers](https://awesome-repositories.com/f/artificial-intelligence-ml/computer-vision-systems/computer-vision/object-detection-tracking/object-tracking-systems/detection-to-track-association/detector-agnostic-trackers.md) — Accepts detection results from any object detector and feeds them into the association algorithm for tracking. - [Detector Integrations](https://awesome-repositories.com/f/artificial-intelligence-ml/computer-vision-systems/computer-vision/object-detection-tracking/object-tracking-systems/detection-to-track-association/detector-integrations.md) — Accepts detection results from any object detector and feeds them into the tracking association algorithm. ([source](https://cdn.jsdelivr.net/gh/foundationvision/bytetrack@main/README.md)) - [Occlusion Recovery](https://awesome-repositories.com/f/artificial-intelligence-ml/dense-pixel-tracking/occlusion-recovery.md) — Recovers occluded objects by re-matching low-confidence detections with unmatched tracklets. - [Detection-to-Tracklet Association Algorithms](https://awesome-repositories.com/f/artificial-intelligence-ml/detection-to-tracklet-association-algorithms.md) — Implements the core ByteTrack association algorithm that matches all detection boxes to tracklets. ([source](https://cdn.jsdelivr.net/gh/foundationvision/bytetrack@main/README.md)) - [Hungarian Matching Losses](https://awesome-repositories.com/f/artificial-intelligence-ml/machine-learning/architectures/neural-network-components/loss-functions/perceptual-loss/content-loss-calculators/focal-loss-calculators/detection-loss-calculators/hungarian-matching-losses.md) — Uses Hungarian algorithm bipartite matching to assign detections to tracklets with minimal cost. - [Object Tracking Frameworks](https://awesome-repositories.com/f/artificial-intelligence-ml/object-tracking-frameworks.md) — Associates every detection box across video frames to produce consistent object identities, including low-confidence ones. - [Linear Motion Models](https://awesome-repositories.com/f/artificial-intelligence-ml/object-tracking/linear-motion-models.md) — Employs Kalman filter linear motion models to predict object positions between video frames. - [Tracking Model Fine-Tuning Pipelines](https://awesome-repositories.com/f/artificial-intelligence-ml/custom-model-training/tracking-model-fine-tuning-pipelines.md) — Provides a pipeline to fine-tune pretrained detectors on custom multi-object tracking datasets. - [Dual Detection Thresholds](https://awesome-repositories.com/f/artificial-intelligence-ml/face-detection/confidence-filtering/prediction-thresholds/dual-detection-thresholds.md) — Ships a dual-threshold strategy that processes both high- and low-confidence detections for robust tracking. - [ONNX Model Exporters](https://awesome-repositories.com/f/artificial-intelligence-ml/machine-learning/infrastructure/model-deployment-and-serving/serialization-and-export-formats/onnx-model-exporters.md) — Exports trained tracking models to the ONNX format for cross-platform inference deployment. - [Tracking Model ONNX Exporters](https://awesome-repositories.com/f/artificial-intelligence-ml/machine-learning/infrastructure/model-deployment-and-serving/serialization-and-export-formats/onnx-model-exporters/tracking-model-onnx-exporters.md) — Exports tracking models to ONNX, TensorRT, ncnn, and DeepStream formats for production deployment. - [Tracking Model TensorRT Optimizations](https://awesome-repositories.com/f/artificial-intelligence-ml/ml-library-integrations/tensorrt-framework-integrations/tracking-model-tensorrt-optimizations.md) — Compiles tracking models into TensorRT engines with FP16 quantization for low-latency GPU inference. - [Tracking Model Fine-Tuning Pipelines](https://awesome-repositories.com/f/artificial-intelligence-ml/model-training-frameworks/vision-model-training/vision-language-training/vision-language-fine-tunings/streamlined-fine-tuning-pipelines/tracking-model-fine-tuning-pipelines.md) — Provides a pipeline to fine-tune pretrained detectors on custom multi-object tracking datasets. ([source](https://cdn.jsdelivr.net/gh/foundationvision/bytetrack@main/README.md)) - [Tracklet Lifecycle Managers](https://awesome-repositories.com/f/artificial-intelligence-ml/tracklet-lifecycle-managers.md) — Manages the full lifecycle of tracklets including creation, update, and deletion based on association results. ### Graphics & Multimedia - [Video Frame Processing](https://awesome-repositories.com/f/graphics-multimedia/video-frame-processing.md) — Processes video files frame by frame to output bounding boxes and tracked object identities. ([source](https://cdn.jsdelivr.net/gh/foundationvision/bytetrack@main/README.md)) ### DevOps & Infrastructure - [Model Export Formats](https://awesome-repositories.com/f/devops-infrastructure/deployment-management/model-export-formats.md) — Exports tracking models to ONNX, TensorRT, ncnn, and DeepStream formats for production inference. ([source](https://cdn.jsdelivr.net/gh/foundationvision/bytetrack@main/README.md))