An MIT-licensed computer vision library for training, running, and exporting a wide range of detection, segmentation, pose, depth, and OCR models.