ArtFormer: Controllable Generation of Diverse 3D Articulated Objects

Environment

conda env create -f env.yml

Data Preprocess

donwload the dataset of PartNet-Mobility unzip it in data/datasets/0_raw_dataset. Then excute the scripts in data/process_data_script one by one.

Train

1. Training of SDF:

python 1_train_SDF.py -c configs/1_SDF/train.yaml

2. Training of Diffusion:

python 2_train_diff.py -c configs/2_Diff/train.yaml

3. Training of Articulation Transformer:

For text condition:

python 3_train_trans.py -c configs/3_TF-Diff/text-train.yaml

For image condition:

python 3_train_trans.py -c configs/3_TF-Diff/image-train.yaml

Evaluation

python 3_pred_trans.py -c configs/3_TF-Diff/<THE-ITEM-YOU-WANT-TO-EVALUATION>.yaml