Nuscenes implementation

[pigtigger]

Hey @zye1996 , this repo is a great work ! and i m wondering if it is convenient for you to update some guidance to train 3dssd on Nuscenes, as i find some nuscenes configs already in tools/cfgs. is it on your roadmap?

[zye1996]

Hey @zye1996 , this repo is a great work ! and i m wondering if it is convenient for you to update some guidance to train 3dssd on Nuscenes, as i find some nuscenes configs already in tools/cfgs. is it on your roadmap?

From the current training result, I cannot reproduce the performance stated by the paper (I get NDS 0.263). It would be great if could try use the configuration to train by yourself and compare the result.

[sixcomicpen]

Dear author, how can I solve the following problem： Your compiler (c++) is not compatible with the compiler Pytorch was built with for this platform, which is g++ on linux.

[markwave]

Hey @zye1996 , this repo is a great work ! and i m wondering if it is convenient for you to update some guidance to train 3dssd on Nuscenes, as i find some nuscenes configs already in tools/cfgs. is it on your roadmap?

hi,i want to train the nuscens_mini dataset on the 3dssd_diou.yaml on the cfg,but it can runs well ,it will occur the error like this below,how can i fix this? Error message:"Traceback (most recent call last): File "train.py", line 208, in main() File "train.py", line 67, in main args, cfg = parse_config() File "train.py", line 56, in parse_config cfg_from_yaml_file(args.cfg_file, cfg) File "/data/cai/3DSSD/pcdet/config.py", line 78, in cfg_from_yaml_file merge_new_config(config=config, new_config=new_config) File "/data/cai/3DSSD/pcdet/config.py", line 52, in merge_new_config if 'BASE_CONFIG' in new_config: TypeError: argument of type 'NoneType' is not iterable "