- Detects 2D pose of multiple people in the image.
- Jointly trained to find the body parts and associations of the body parts with the individuals.
- Earlier approaches were based on finding humans in the image, so depended on the number of persons in the image.
- Also, if the human detector failed, everything fell apart.
The system takes, as input, a color image of size w × h (Fig. 2a) and produces, as output, the 2D locations of anatomical keypoints for each person in the image.
The model is split into two branches: one predicts the confidence maps for the body parts and the second branch predicts the affinity between those body parts.