Калибровка камер по видео через AutoMagicCalib

★ 7.6 · content

amc-run-video-calibration is a Claude Code skill that calibrates a multi-camera dataset from pre-recorded MP4 files via the AutoMagicCalib REST API — no CLI scripts or Docker bind-mounts required, just a running AMC microservice. It creates a project, uploads videos sorted alphabetically (the server assigns camera indices by upload order), and auto-scans the video directory and its parent for calibration settings (`settings.json`), alignment data (`alignment_data.json`), and a layout image (`layout.png`). Optional inputs include a ground-truth ZIP for evaluation metrics, per-camera focal lengths, and detector selection — `resnet` (fast, default) or `transformer` (better under occlusion). Designed for computer vision engineers and developers who have locally stored, time-synchronized camera recordings and a running AMC stack; live RTSP streams are handled by a separate skill.