search

PRBonn / SHINE_mapping

🌟 SHINE-Mapping: Large-Scale 3D Mapping Using Sparse Hierarchical Implicit Neural Representations (ICRA 2023)
MIT License
447 stars 32 forks source link

bad result for maicity with voxel size 20cm #2

Closed JunyuanDeng closed 1 year ago

JunyuanDeng commented 1 year ago

Thanks for your excellent work!

But when I test the mappings for maicity with voxel size 20 cm, the result turns to like below:

image

It seems there are many holes and the disappear of pavement.

I just simply change the hyperparameter leaf_vox_size to 0.2 in maicity_incre_reg.yaml . How should I modify it to get a good result?

YuePanEdward commented 1 year ago

Thanks for your interest in our work. The issue may be caused by the synthetic dataset and the marching cubes library we used. You may set the mc_res_m: 0.1 and you are expected to get a better result without the holes.

We will try to fix this marching cubes issue ASAP

JunyuanDeng commented 1 year ago

Thanks for your reply, I will test it later!

JunyuanDeng commented 1 year ago

I have tried with mc_res_m: 0.1, it has no stranges holes anymore, but the result seems to be less good... 微信截图_20230114005830

aartykov commented 1 year ago

I could not reproduce the results in table 2 with the MaiCit dataset too. What should I do to get similar results as the ones mentioned on the paper? Thanks in advance.

YuePanEdward commented 1 year ago

I could not reproduce the results in table 2 with the MaiCit dataset too. What should I do to get similar results as the ones mentioned on the paper? Thanks in advance.

Thanks for your interest in our work. You may check a similar issue here https://github.com/PRBonn/SHINE_mapping/issues/3#issuecomment-1367416983.