pytracking

Why we included this project

Researchers and engineers working on visual object tracking and video object segmentation will find one of the more complete codebases here. It ships official implementations of established trackers like DiMP, PrDiMP, ToMP, and RTS, and pairs them with the datasets, evaluation scripts, and building blocks needed to reproduce published results and try your own ideas. The bundled LTR framework covers the heavy parts of training tracking networks, including data sampling and network modules, so you can focus on the model itself. A model zoo provides pretrained weights and benchmark numbers, which makes it easy to get a baseline for comparison. Expect a research-oriented toolkit rather than a plug-and-play product; you'll likely need to integrate pieces into your own pipeline.