run_subpixelmaxima

lightning_pose.models.heads.heatmap.run_subpixelmaxima(heatmaps: Tensor, {'__torchtyping__': True, 'details': ('batch', 'num_keypoints', 'heatmap_height', 'heatmap_width'), 'cls_name': 'TensorType'}], downsample_factor: int, temperature: tensor) → Tensor, {'__torchtyping__': True, 'details': ('batch', 'num_keypoints',), 'cls_name': 'TensorType'}]]

Use soft argmax on heatmaps.

Parameters:

• heatmaps – output of upsampling layers

• downsample_factor – controls how many times upsampling needs to be performed

• temperature – temperature parameter of softmax; higher leads to tighter peaks

Returns:

tuple

• soft argmax of shape (batch, num_targets)

• confidences of shape (batch, num_keypoints)