Robust Stereo Calibration for Improved 2D-3D Projection in Real-World Pose Estimation - Experiment Image Data

This dataset comprises raw image data collected during our research project titled "Robust Stereo Calibration for Enhanced 2D-3D Projection in Real-World Pose Estimation." The study investigates novel approaches to improve stereo calibration for precise 3D projection and 2D back-projection in pose estimation.The dataset includes all image frames used for performing intrinsic and extrinsic calibration using the open-source Python library, OpenCV. We analyze various factors influencing calibration quality, such as camera positioning, calibration board sizes, and capture methods.The dataset structure is organized as follows: "frames/[<i>experiment_group</i>]/[<i>experiment_group</i>]_[<i>frame_num</i>].jpg". Here, <i>experiment_group</i> follows the format "[<i>angle</i>]_[<i>size</i>]_[<i>camera</i>]_[<i>group</i>]":<i>Angle</i> denotes the external experimental setting, representing the camera pair's viewing angle: close (15°), medium (45°), or wide (90°).<i>Size</i> indicates the calibration board size used in the experiment: A1, A2, or A3.<i>Camera</i> identifies the specific camera used for data capture: C1 (primary), C2, or C3 (pair).<i>Group</i> specifies the stereo camera pairing: G12 or G13, indicating simultaneous data capture from different perspectives.If the group tag is absent and "_internal" replaces it, the frames are used for intrinsic camera parameter estimation. Frames suffixed with "_single" indicate data captured using a still-capture image method; frames without this suffix were extracted from video footage captured using a video-capture method.