For Nuscenes dataset, when we get the projection matrix in ptq/export_onnx.py line 54,
intrinsic = torch.tensor(img_meta["lidar2img"]["intrinsic"][:3, :3])
The intrinsic output is [[1 0 0 0][0 1 0 0][0 0 1 0][0 0 0 1]], which is not the camera intrinsic.
The real camera intrinsic is in ['lidar2img_extra']['cam_intrinsic'].
So which is right intrinsic?
For Nuscenes dataset, when we get the projection matrix in ptq/export_onnx.py line 54, intrinsic = torch.tensor(img_meta["lidar2img"]["intrinsic"][:3, :3]) The intrinsic output is [[1 0 0 0][0 1 0 0][0 0 1 0][0 0 0 1]], which is not the camera intrinsic. The real camera intrinsic is in ['lidar2img_extra']['cam_intrinsic']. So which is right intrinsic?