Python Wrapper for the Azure Kinect Recorder (Camera model)

Hello

where can I find the python wrapper for the azure kinect especially for the recording with the azure kinect? I can't find it. I need to know which camera model is used for the recording to transform the point clouds with the right matrix and to set the right intrinsic parameters. Because if I use the PinholeIntrinsicParameters provided my point clouds are heavily distorted. Although I record the data with the azure kinect via azure_kinect_recorder.py. Afterwards I generate point clouds out of the mkv file with this script:
```
def generate_point_cloud_from_color_and_depth_images(self, pcd_dir):
for color, depth in zip(self.color_file_list, self.depth_file_list):
filename, ending = self.color_file_list[self.idx].split(".")
color_raw = o3d.io.read_image(self.color_dir + color)
depth_raw = o3d.io.read_image(self.depth_dir + depth)

rgbd_image = o3d.geometry.RGBDImage.create_from_color_and_depth(color=color_raw, depth=depth_raw, depth_scale=1000.0, convert_rgb_to_intensity=False)
# intrinsic = o3d.camera.PinholeCameraIntrinsic(1280, 720, 601.1693115234375, 600.85931396484375, 637.83624267578125, 363.8018798828125) this intrinsic deliever the best result but it is still not useful
point_cloud = o3d.geometry.PointCloud.create_from_rgbd_image(rgbd_image, o3d.camera.PinholeCameraIntrinsic(o3d.camera.PinholeCameraIntrinsicParameters.Kinect2ColorCameraDefault))
point_cloud.transform([[1, 0, 0, 0], [0, -1, 0, 0], [0, 0, -1, 0], [0, 0, 0, 1]])
point_cloud_filename = "{0}\\pcd_1\\{1}.pcd".format(pcd_dir, filename)
o3d.io.write_point_cloud(point_cloud_filename, point_cloud)
print("Writing to {}".format(point_cloud_filename))
self.idx = self.idx + 1
```
I already tested all PinholeCameraIntrinsicParameters provided but the point clouds are heavily distorted or the depth isn't represented right.
The intrinsic_matrix in the instrinsic.json is as follows:
```
[
613.01324462890625,
0.0,
0.0,
0.0,
613.0467529296875,
0.0,
639.76904296875,
369.3740234375,
1.0
]
```
How can I figure out which camera model is used for recording without the python wrapper? Does it use a pinhole model or an rational polynomial model?

Thank you in advance.