PartDistill

[Project Page] [Code] [Arxiv] [Video] [Poster]

This is the official repository of our CVPR 2024 paper, "PartDistill: 3D Shape Part Segmentation by Vision-Language Model Distillation".

Data and required 3D backbone checkpoint (Point-M2AE)

  • The prapared data can be downloaded here. The content details are explained in data_format.md. After downloading, please extract the zip file.
  • Point-M2AE checkpoint can be download here

Train

To train the model, please use script bellow

CUDA_VISIBLE_DEVICES=<gpu_idx> python train.py \
 --data_path<dir_path_storing_preprocessed_data> \
 --ckpt_dir=<dir_to_store_ckpt> \
 --backbone_path=<path_to_pointm2ae_ckpt> \
 --lr=0.003 \
 --batch_size=16 \
 --category chair knife \
 --n_epoch=25 \
 --exp_suffix=exp1

For example:

CUDA_VISIBLE_DEVICES=8 python train.py \
 --data_path=/disk2/aumam/dataset/shapenet/shapenet_preprocessed_partdistill \
 --ckpt_dir=/disk2/aumam/dev/multimodal_distillation/checkpoints \
 --backbone_path=/disk2/aumam/dev/multimodal_distillation/checkpoints/point_m2ae_pre-train.pth \
 --lr=0.003 \
 --batch_size=16 \
 --category chair knife \
 --n_epoch=25 \
 --exp_suffix=exp1

Note

Install this library to use Point-M2AE

pip install --upgrade https://github.com/unlimblue/KNN_CUDA/releases/download/0.2/KNN_CUDA-0.2-py3-none-any.whl

Cite

@inproceedings{umam2023partdistill,
  title = {PartDistill: 3D Shape Part Segmentation by Vision-Language Model Distillation},
  author = {Umam, Ardian and Yang, Cheng-Kun and Chen, Min-Hung and Chuang, Jen-Hui and Lin, Yen-Yu},
  booktitle = {IEEE/CVF International Conference on Computer Vision (CVPR)},
  year = {2024},
}