Topic · Tasks

Keypoints and pose

Everything on this site about keypoints and pose, in the order to read it: a page is listed after the pages it builds on. The figures below are the ones those pages use to explain it; each is live: move a slider and it recomputes.

In reading order

  1. Paper · read first: Object Detection — A Quick Read, Understanding Focal Loss —A Quick Read
  2. Paper · read first: Triplet Loss Demystified in 5 mins

The figures these pages use

Pulled from the pages above. Move a slider and the figure recomputes, exactly as on its page.

From CenterNet: Objects as Points — A Comprehensive Guide › 2. Proposed Algorithm

1.6
A keypoint heatmap. Darker cells are closer to 1. Training treats the centre as the positive and softens the penalty around it.

From CenterNet: Objects as Points — A Comprehensive Guide › 2. Proposed Algorithm

0.64.0
Loss at one heatmap cell. Left: the centre cell wants Ŷ → 1. Right: a neighbouring cell wants Ŷ → 0, but the penalty shrinks as its Gaussian value Y approaches 1.

From CenterNet: Objects as Points — A Comprehensive Guide › 2. Proposed Algorithm

4
The rounding error the offset head recovers. Left: the fraction lost when a pixel coordinate is mapped to the low-resolution map. Right: the L1 loss on it is flat-sloped, so every error counts the same.

From CenterNet: Objects as Points — A Comprehensive Guide › 2. Proposed Algorithm

0.11.0
Weighting the three losses. Raw losses (left of each pair) and what each contributes to L_det after weighting. The paper uses λ_size = 0.1 and λ_off = 1.

Tags that go with it

Anchor-free detection 1Object detection 1Re-identification 1Traffic analytics 1