Main features of OpenTL
- Modularized, object-oriented software architecture
- Common abstractions inside each layer: Bayesian filters, visual modalities, pose representations, data storage, image processing etc.
- Realtime performance: optimized algorithms, with support for multi-core and GPU processing
- Different Bayesian filters, including particle filters (SIR, MCMC) as well as Kalman-based filters (EKF, IF)
- A large variety of visual modalities: for example color statistics, background subtraction, Contours, local keypoints, intensity gradients, templates, blobs, etc.
- Robust improvements: data fusion (static, dynamic), models combining online and offline data, color space conversions, etc.
- Articulated objects
- General-purpose XML parser for 3D models
- Support for different camera types: USB webcams, FireWire cameras, GigE Vision-based cameras
- Multiple platforms: currently Microsoft Windows (XP, Vista) and Linux (Ubuntu Hardy)
- Generic sensor abstraction: camera, lidar, radar, gps, video file
- Combining multi-camera, multi-target and multi-modality support