AquaCal Example Dataset: Real Multi-Camera Underwater Rig

This calibration dataset supports the AquaCal refractive multi-camera calibration library. It contains synchronized sequences from a 13-camera array (12 standard plus 1 fisheye) capturing ChArUco calibration boards both in air and through a flat air-water interface. Version 2.0.0 replaces the frame-subsampled extraction published in version 1.0.0 with the complete frameset. Version 1.0.0 carried 60 extrinsic frames per camera and reproduced materially different numbers from those reported in the AquaCal paper; this version carries the full 262 and reproduces Section 3 directly. Dataset Contents:- 3,406 extrinsic calibration frames (13 cameras x 262 frames), lossless PNG- 561 intrinsic calibration frames (13 cameras, variable counts per camera by design, ranging from 31 to 116)- Two AquaCal configuration files: config_paper.yaml, which reproduces Section 3 of the paper, and config_quickstart_not_paper.yaml, a faster first-contact run that explicitly does not- reference_outputs/, the reference calibration and diagnostics produced by running config_paper.yaml over these frames, plus the residual and error tables used in the paper's analysis- reference_calibration.json at the archive root, for AquaCal's load_example() Frames are every 30th frame of the source recordings, which is the subsampling the published calibration run used, so this archive carries the frames that produced the published numbers rather than a re-sampling of them. Reference results were produced with AquaCal 1.8.0. Reported values include a mean primary reprojection error of 0.824 px, an inter-corner mean absolute error of 0.258 mm on 60 mm nominal spacing across 7,762 comparisons, and a recovered water surface at z = 1.074 m. Camera Hardware: WhiteMatter e3Vision system Calibration Boards:- 12x9 ChArUco (60mm squares, 45mm markers) for extrinsics- 11x8 ChArUco (20mm squares, 15mm markers, legacy pattern) for intrinsics