raw input about f-pointnet

[Veincore]

Can it be applied to real-time point cloud and image input?If I want to implement this feature, what should I do.Expect your reply.

It takes about 300ms per frame.
We used raw data of KITTI dataset to test our code. Change this directory to your dataset ( infer/prep_data_infer.py , line 487 ). Make sure the KITTI data is organized as required in line 208.
In addition, we used gluoncv for YOLO. So mxnet should be installed.

[zhouzhou33]

@Veincore Thank you very much for your reply.What if I want to use the point cloud data from the sensor directly?

[Veincore]

@Veincore Thank you very much for your reply.What if I want to use the point cloud data from the sensor directly?

Change function get_lidar ( infer/prep_data_infer.py , line 225), return data from your sensor. Weights of pointnets should be retrained.

[zhouzhou33]

@Veincore Thank you very much indeed for your answer.

[sbastani]

Thanks for the good work!
Have you also done this for SUNRGBD, too? Otherwise, could you please guide me how to extend this to support inference with sunrgbd or directly from rgb-d sensor?

[zhouzhou33]

Sorry, No. I've been doing something else before. Now I'm back in touch with this project. Maybe you solved this problem?
@sbastani

[PauloMendes33]

zhouzhou33 have you been able to use your own data and sensor?