Process

Object detection and segmentation

You might say: recognising objects

An image model predicts object classes and locations; segmentation predicts the pixels belonging to each object.

What it does

Provides targets for grasping and obstacles for avoidance.

Ask

What are precision, recall, and AP on images from this robot and lighting?

Typical

COCO's primary AP averages detections over IoU thresholds from 0.50 through 0.95.

What to look for

1. Object classes and locations on an image
2. Pixels belonging to each object
3. Grasp targets and avoidance obstacles

Parts

• Precision and recall: Precision is the share of detections that are correct.

Sources

1. COCO Consortium, COCO Detection Evaluation, Detection EvaluationFiftyOne provides COCO-style object detection evaluation to not only compute AP but also easily visualize individual sample- and object-level results.
2. Russ Tedrake, Robotic Manipulation, Ch. 10, Object Detection and SegmentationObject detection and segmentation There is a lot to know about modern object detection and segmentation pipelines.

Use it on your project

teardown-apply: hold "Object detection and segmentation" (object-detection) from the teardown at https://teardownatlas.com/robotics/ against my project

Paste it into a coding agent that has the teardown skills.

An AI agent wrote this map with the teardown skill. It can contain errors: check the sources before you rely on a part. About and legal