ultralytics-yolo

What problem does it solve?

Simplifies the end-to-end workflow of training, validating, running inference, exporting, and deploying Ultralytics YOLO models so practitioners can move from raw images to production-ready models without guesswork.

Core Features & Use Cases

• Model training & validation: Train YOLOv8/26/11 models on custom datasets using data.yaml, multi-GPU, or Apple MPS and monitor metrics like mAP, precision, and recall.
• Inference & tracking: Run predictions on images, videos, streams, and directories, and perform multi-object tracking with BoT-SORT or ByteTrack.
• Export & deployment: Export to ONNX, TensorRT, CoreML, TFLite, OpenVINO and support quantization flags (FP16/INT8) for edge and server deployment.
• Use case: Prepare a custom detection dataset, train a balanced yolo26s model, validate its mAP, export to ONNX, and run batched inference for analytics and tracking.

Quick Start

Train a yolo26n model on my dataset using the data.yaml at path /path/to/data.yaml for 100 epochs and export the best checkpoint to ONNX.