Extrinsic stereo camera calibration using human pose recognition

Extrinsic camera calibration is essential for any computer vision tasks in a multi-camera setup. Installing cameras in the field can be costly and troublesome, especially when even a single camera is moved and a new calibration is needed. The thesis proposal consists in deepening a completely automatic approach enabling non-expert in a fast and accurate calibration using pedestrian detection and/or human pose recognition instead of calibration objects.