GitHub - N0vel/weighted-hausdorff-distance-tensorflow-keras-loss: Weighted Hausdorff Distance Loss: use it as point cloud similarity metric based loss for keras and tf. Useful in keypoint detection.

Weighted Hausdorff Distance Loss: use it as a point cloud similarity metric based loss for keras and tf. This loss requires a huge tensor with dimensions (number_of_pixels * number_of_keypoints if I remember correctly) of float values. So high res picture with thousands of keypoints will consume A LOT of GPU memory (at least 1 GB for 512 pixels x 512 pixels x 1000 keypoints with float32 type). It doesn't matter if you want to detect only several points in an image.