# hrnet

Python API: `depthai_nodes.node.parsers.utils.hrnet`

## Functions

### compute_hrnet_keypoints

```python
def compute_hrnet_keypoints(heatmaps: np.ndarray) -> tuple[np.ndarray, np.ndarray]:
```

Find each HRNet heatmap maximum and normalize its coordinates.

Parameters

 * `heatmaps` (`np.ndarray`): Heatmaps of shape `(1, keypoints, height, width)` or `(keypoints, height, width)`. The first axis is
   removed when its size is 1.

Returns

 * `tuple[np.ndarray, np.ndarray]`: A pair of normalized `(N, 2)` XY coordinates and `(N,)` peak scores clipped to [0, 1].

Raises

 * `ValueError`: If the heatmaps are not 3D after removal of a singleton first axis.