dgcnn_rsmix

RSMix for DGCNN(PyTorch)

We utilize the original released codes of DGCNN, which is implemented with PyTorch.

• RSMix is implemented in rsmix_provider.py.

Prepare Dataset

• You can get sampled point clouds of ModelNet40 (XYZ and normal from mesh, 10k points per shape) here (1.6GB).

Move the uncompressed data folder or create symbolic link to data/modelnet40_normal_resampled.

• You can also get sampled point clouds of ModelNet40 (XYZ and normal from mesh, 10k points per shape) as hdf5 format here (435MB).

Move the uncompressed data folder or create symbolic link to data/modelnet40_ply_hdf5_2048.

Environment

Follow the environment setting of original DGCNN code.

[DGCNN PyTorch]

Point Cloud Classification

Training

Run the training script(epoch=500):

• 1024 points

python main.py --exp_name=rsmix_dgcnn_1024 --model=dgcnn --num_points=1024 --k=20 --use_sgd=True --beta 1.0 --epochs 500

• 2048 points

python main.py --exp_name=rsmix_dgcnn_2048 --model=dgcnn --num_points=2048 --k=40 --use_sgd=True --beta 1.0 --epochs 500

Note: if you want to test the combinations of augmentations with RSMix,

you can selectively input the augmentation-related arguments from one of the follow arguments.

• conventional data agumentation arguments:

--shuffle : Random shuffle augmentation

--jitter : Jitter augmentation

--rot : Random Rotation augmentation

--rdscale : Random Scaling augmentation

--shift : Random Shif augmentation

• RandDrop augmentation argument:

--rddrop : RandDrop augmentation

Additionally, if you want to test with ModelNet10, please input the argument --modelnet10. Default dataset is ModelNet40.

Evaluation

Run the evaluation script after training finished:

• 1024 points

python main.py --exp_name=rsmix_dgcnn_1024_eval --model=dgcnn --num_points=1024 --k=20 --use_sgd=True --eval=True --model_path=checkpoints/rsmix_dgcnn_1024/models/model.t7

• 2048 points

python main.py --exp_name=rsmix_dgcnn_2048_eval --model=dgcnn --num_points=2048 --k=40 --use_sgd=True --eval=True --model_path=checkpoints/rsmix_dgcnn_2048/models/model.t7

Evaluation with pretrained model

Run the evaluation script with pretrained models:

• 1024 points

python main.py --exp_name=rsmix_dgcnn_1024_eval --model=dgcnn --num_points=1024 --k=20 --use_sgd=True --eval=True --model_path=pretrained/model.1024.t7

• 2048 points

python main.py --exp_name=rsmix_dgcnn_2048_eval --model=dgcnn --num_points=2048 --k=40 --use_sgd=True --eval=True --model_path=pretrained/model.2048.t7