search

WeijingShi / Point-GNN

Point-GNN: Graph Neural Network for 3D Object Detection in a Point Cloud, CVPR 2020.
MIT License
540 stars 112 forks source link

Questions regarding the naming of files #2

Closed dirk61 closed 4 years ago

dirk61 commented 4 years ago

First of all, I really like the work you've done! Good job!😀

I'm quite confused about the "T0"~"T3" and the combination of "train" and "eval" in some of the file names. For example, in the configs/ directory, what's the difference between car_auto_T0_train_config & car_auto_T3_train_config & car_auto_T3_train_train_config? It would be really helpfu if I can understand how you name these files!😀

2020315-91139
WeijingShi commented 4 years ago

The checkpoints are named after the ablation study we did in the paper.

In general, the name follows [Class type] + [Use auto-registration (auto) or not (fixed)] + [Number of layers] + [data split file] + [config/train_config/eval_config].

For example, _car_auto_T3_train_trainconfig means: a train_config file, using T=3 layers, each layer use auto-registration, trained on the KITTI train split.

I hope this helps to clear the confusion.