Without lidar about radar_depth HOT 5 CLOSED

Hi @zazgf,

It depends on the sparsifier argument passed to the dataset construction (the name itself is a bit confusing). So if the sparsifier is "radar", then the inputs contain the concatenation of images and raw radar depth maps. If it's "radar_filtered", then the inputs are composed by images and filtered radar depth maps instead.

Therefore, as you can see in README.md, the sparsifier is set to "radar".

from radar_depth.

With reading the program, I learned a lot, Thank you very much for releasing the code.

from radar_depth.

Hi @zazgf ,

The part you mentioned is actually in the data loader and that part was used to generate different sparse patterns for the experiments in Table 2 from our paper. So you can generate (1) normal radar measurements, (2) radar points filtered by the ground truth LiDAR scans, or (3) radar points filtered by the trained classifyer, etc. In the main experiment, we use (1), so there should be no concern related to "using lidar data to do the filtering".

from radar_depth.

@brade31919 Thanks for your reply.
Does this mean that I can achieve the project function without lidar?
Thanks again.

from radar_depth.

@brade31919 Get your point, I didn't realize this before.I need pay more attention on code reading. Thanks.

from radar_depth.