Gesto

Gesto is an open framework designed for gesture and pose recognition, offering a complete pipeline from data capture to model deployment. Key features: * Capture gesture datasets from webcam * Train custom recognition models * Detect gestures in real-time * Supports static poses and motion sequences * Entirely offline operation The framework consists of two main components: Gesto Labeller (Community Edition) and the gesto Python package. Gesto Labeller is a user-friendly desktop application that enables users to record and label gesture samples from their webcam. It supports capturing hands, two hands, full-body poses, and specific limb movements. Users can explore, relabel, and trim samples visually, with all data exported as plain .npy files, ensuring data ownership and privacy. This component is ideal for creating custom datasets without reliance on cloud services or external labelling platforms. Once datasets are prepared, the gesto Python package takes over, allowing users to train and run recognition models. With simple command-line operations like `gesto train` and `gesto detect`, users can develop models for static or sequence-based recognition and deploy them live on a webcam or video file. The package ensures consistent landmark extraction and normalization across all stages, from capture to detection, for accurate model performance. Gesto is perfect for researchers, developers, and hobbyists looking to build custom gesture-controlled applications, interactive experiences, or conduct behavioral analysis without complex setups or reliance on external cloud infrastructure. It offers a powerful, accessible way to integrate pose and gesture understanding into various projects.