SAF is introduced to optimize predicted landmark detection heatmaps by enforcing an optimum mean squared error (MSE) between a pair of predicted heatmaps. It also enforces a secondary optimum MSE between a predicted heatmap and a zero matrix.
The formula of SAF is:
The figure of SAF without the Min term is:
Configuring SAF requires following steps:
- Determine the MSE between a pair of ground truth heatmap.
- Determine the value of coefficient a which determines the slope of SAF around the minimum point in Quadrant 1 of the Cartesian coordinate system.
- Compute coefficient b using equation:
$$b = \frac{1}{a\times Optimum MSE^2}$$ - Compute coefficient c using equation:
$$c = \frac{\log(Min)}{\log(\frac{ax^2+1}{bx})}, x = 0.0061$$
Citation: TBD