search

PRBonn / pole-localization

Online Range Image-based Pole Extractor for Long-term LiDAR Localization in Urban Environments
MIT License
169 stars 22 forks source link

How do you do the inference with this with a pretrained model? #10

Closed jediofgever closed 1 month ago

donghao51 commented 1 month ago

please refer to ncltpoles_learning.py

jediofgever commented 1 month ago

Thank you

jediofgever commented 1 month ago

What I have is basically only a LIDAR point cloud, which I could convert to a range image with your code. I see that a "session" requires a lot more fields, given in nclt dataset, if we imagine I only have .npy file with my point cloud, what would be the most starigh forward way to inference on this?

donghao51 commented 1 month ago

you can simply convert .npy to range image, normalize it, and use this function https://github.com/PRBonn/pole-localization/blob/09870584b2d35e8ce426f4c5560f81a355c8892b/src/poles_extractor.py#L6

jediofgever commented 1 month ago

what did you mean by normalize it?

donghao51 commented 1 month ago

it is already normalized in the code so you don't need to do this https://github.com/PRBonn/pole-localization/blob/09870584b2d35e8ce426f4c5560f81a355c8892b/src/poles_extractor.py#L22