YOLO: Decoding the Box Centre

Intermediate

The problem, from first principles

YOLO predicts a box centre relative to the cell that owns it. A sigmoid keeps the predicted offset between zero and one, so the centre cannot leave its cell, and the offset is converted back to image pixels.

From theory to code

Implement decode_xy(sx, sy, col, row, S, img), returning the box centre in pixels.

Constraints

Use a sigmoid on the raw offsets.

Hints

Hint 1

Apply the sigmoid to each raw offset, add the cell index, then multiply by the cell size in pixels.

Keywords

research-paperscomputer-visiondetectionobject-detection

Python

Run your code to see results

No results yet

Run your code to check it against the examples, or submit to run every test.