OpenPose

OpenPose represents a landmark in computer vision, being the first real-time multi-person system to jointly detect human body, hand, facial, and foot keypoints (totaling 135 keypoints) on single images. Developed at Carnegie Mellon University (CMU), the architecture utilizes a bottom-up approach through Part Affinity Fields (PAFs) to associate body parts with individuals, significantly outperforming top-down methods in crowded environments. In the 2026 landscape, while newer transformer-based models exist, OpenPose remains the benchmark for low-latency C++ deployments and research validation. Its technical stack is rooted in Caffe and OpenCV, with robust wrappers for Python and Unity, making it highly versatile for industries ranging from sports science to high-fidelity character animation. The system is particularly valued for its 'bottom-up' inference, which maintains constant processing speed regardless of the number of people in the frame, a critical requirement for real-time surveillance and large-scale public motion tracking. Commercial adoption requires a specific license from CMU, ensuring it remains a staple in professional-grade biometric solutions.