DeepLabCut is an open-source toolbox for 3D markerless pose estimation across species and behaviors. It uses a deep neural network trained on limited data to accurately track user-defined body parts. The toolbox includes a Python package with graphical user interfaces and active-learning based network refinement. It provides a step-by-step guide for use, allowing researchers to estimate subject poses efficiently. DeepLabCut is versatile, requiring minimal labeled data and enabling real-time processing. It is suitable for a wide range of applications, including behavioral analysis in labs, sports, gait analysis, and medicine. The toolbox is robust in dynamic environments with varying backgrounds and can be used with multiple cameras for 3D pose estimation. It is compatible with various hardware and software, including GPUs and different video formats. DeepLabCut has been applied to various organisms, including mice, zebrafish, flies, babies, and cheetahs. It is compared to other pose-estimation methods, showing advantages in speed, accuracy, and adaptability. The toolbox is open-source, free, and provides interactive tools for labeling and refining data. It is designed to be user-friendly, with a step-by-step procedure for creating projects, labeling data, and training networks. The toolbox is efficient for high-throughput video analysis and can be used on standard computers with a compromise on speed. It is compatible with various operating systems and hardware, including consumer-grade GPUs. The toolbox is installed via Python and TensorFlow, with options for GPU or CPU support. It includes functions for creating projects, selecting data, labeling frames, checking annotations, creating training datasets, and training networks. The toolbox is suitable for a wide range of applications and is designed to be efficient and accurate for pose estimation in various scenarios.DeepLabCut is an open-source toolbox for 3D markerless pose estimation across species and behaviors. It uses a deep neural network trained on limited data to accurately track user-defined body parts. The toolbox includes a Python package with graphical user interfaces and active-learning based network refinement. It provides a step-by-step guide for use, allowing researchers to estimate subject poses efficiently. DeepLabCut is versatile, requiring minimal labeled data and enabling real-time processing. It is suitable for a wide range of applications, including behavioral analysis in labs, sports, gait analysis, and medicine. The toolbox is robust in dynamic environments with varying backgrounds and can be used with multiple cameras for 3D pose estimation. It is compatible with various hardware and software, including GPUs and different video formats. DeepLabCut has been applied to various organisms, including mice, zebrafish, flies, babies, and cheetahs. It is compared to other pose-estimation methods, showing advantages in speed, accuracy, and adaptability. The toolbox is open-source, free, and provides interactive tools for labeling and refining data. It is designed to be user-friendly, with a step-by-step procedure for creating projects, labeling data, and training networks. The toolbox is efficient for high-throughput video analysis and can be used on standard computers with a compromise on speed. It is compatible with various operating systems and hardware, including consumer-grade GPUs. The toolbox is installed via Python and TensorFlow, with options for GPU or CPU support. It includes functions for creating projects, selecting data, labeling frames, checking annotations, creating training datasets, and training networks. The toolbox is suitable for a wide range of applications and is designed to be efficient and accurate for pose estimation in various scenarios.