Learn
Guides on training data, annotation & robotics
Practical, no-fluff guides on collecting and labeling data for computer vision and robot learning.
How to build a robot imitation-learning dataset
A practical guide to collecting demonstration data for robot imitation learning: what signals a policy needs (hand pose, the manipulated object, grasp/touch contact, 6DoF), how to capture them, and how to export a clean dataset.
ReadCOCO vs YOLO vs Pascal VOC vs JSONL: which annotation format should you use?
A plain-English comparison of the COCO, YOLO, Pascal VOC and JSONL annotation formats — how each stores boxes and labels, what they can and can't represent, and when to pick each for your training pipeline.
ReadWhat is hand–object contact annotation, and why does it matter?
Hand–object contact annotation records whether and how a hand is interacting with an object — grasp or touch — frame by frame. Here's what it is, how it's computed, and why it's the key signal for robot manipulation datasets.
Read