MinkowskiEngine icon indicating copy to clipboard operation
MinkowskiEngine copied to clipboard
verified publisher icon NVIDIA

Pont cloud density

Open shreyakatari opened this issue 4 years ago • 2 comments

Hi @chrischoy I have trained a model for semantic segmentation on my own data . My data has large halls having so many walls in it when I tested my model on other files output has very less point cloud density walls are coming very thin and even some walls are missing. My model is having 65 as miou. Can you please help me out why points are missing during testing?

shreyakatari avatar Sep 19 '21 15:09 shreyakatari

i think maybe you can get raw resolution by API : https://github.com/NVIDIA/MinkowskiEngine/blob/b975751d89732bf8048418770b349489baca864a/MinkowskiEngine/utils/quantization.py#L192

suyunzzz avatar Dec 23 '21 07:12 suyunzzz

Hello，I also want to train my own network about semantic segmentation,but I can't find any code about this task,what i can find is just the indoor.py which is used to test the trained model. So can you tell me how to train my own network.Thank you very much!!!!!!!!!!!

kunzaiyy avatar Jan 10 '22 03:01 kunzaiyy