Eryck Zhou

A super simple BLOG for Artifical Intelligence.

YOLOv3

29 July 2020

image

Photo by unsplash-logoHuper Earle

Origin: YOLOv3: An Incremental Improvement

Improvement

1. New structure

2. Mutiscale Structure

3 scales and 3 anchors per scale per grid:

  • small scale (13 x 13) ——> large anchor
  • mid scale (26 x 26) ——> medium anchor
  • large scale (52 x 52) ——> small anchor image

3. Change Classfication

  • 80 classes, from softmax ——> logistic

Using a softmax imposes the assumption that each box has exactly one class which is often not the case. A multilabel approach better models the data.

4. Use FPN

image

Summary

Output

  • 13 x 13 x 3 * (4+1 + 80)
  • 26 x 26 x 3 * (4+1 + 80)
  • 52 x 52 x 3 * (4+1 + 80)