Skip to content

Latest commit

 

History

History
86 lines (70 loc) · 2.76 KB

EVALUATE.md

File metadata and controls

86 lines (70 loc) · 2.76 KB

Evaluation Script

Outline

Classification

Architecture

  • DGCNN
python SimpleView/main.py --entry test_corrupt --model-path pretrained_models/DGCNN.pth --exp-config SimpleView/configs/dgcnn_dgcnn_run_1.yaml
  • PointNet
python SimpleView/main.py --entry test_corrupt --model-path pretrained_models/PointNet.pth --exp-config SimpleView/configs/dgcnn_pointnet_run_1.yaml
  • PointNet++
python SimpleView/main.py --entry test_corrupt --model-path pretrained_models/PointNet2.pth --exp-config SimpleView/configs/dgcnn_pointnet2_run_1.yaml
  • RSCNN
python SimpleView/main.py --entry test_corrupt --model-path pretrained_models/RSCNN.pth --exp-config SimpleView/configs/dgcnn_rscnn_run_1.yaml
  • SimpleView
python SimpleView/main.py --entry test_corrupt --model-path pretrained_models/SimpleView.pth --exp-config SimpleView/configs/dgcnn_simpleview_run_1.yaml
  • PCT
python PCT/main.py --exp_name=test --num_points=1024 --use_sgd=True --eval_corrupt=True --model_path pretrained_models/PCT.t7 --test_batch_size 8 --model PCT
  • GDANet
python GDANet/main_cls.py --eval_corrupt=True --model_path pretrained_models/GDANet.t7
  • PAConv
python PAConv/obj_cls/main.py --config PAConv/obj_cls/config/dgcnn_paconv_test.yaml --model_path ../pcdrobustness/pretrained_models/PAConv.t7 --eval_corrupt True
  • CurveNet
python3 CurveNet/core/main_cls.py --exp_name=test --eval_corrupt=True --model_path pretrained_models/CurveNet.t7
  • RPC
python PCT/main.py --exp_name=test --num_points=1024 --use_sgd=True --eval_corrupt=True --model_path pretrained_models/RPC.t7 --test_batch_size 8 --model RPC

Augmentation

  • DGCNN + PointWOLF
python PointWOLF/main.py --exp_name=test --model=dgcnn --num_points=1024 --k=20 --use_sgd=True --eval_corrupt=True --model_path pretrained_models/DGCNN_PointWOLF.t7
  • DGCNN + RSMix
python PointWOLF/main.py --exp_name=test --model=dgcnn --num_points=1024 --k=20 --use_sgd=True --eval_corrupt=True --model_path pretrained_models/DGCNN_RSMix.t7
  • DGCNN + WOLFMix
python PointWOLF/main.py --exp_name=test --model=dgcnn --num_points=1024 --k=20 --use_sgd=True --eval_corrupt=True --model_path pretrained_models/DGCNN_WOLFMix.t7
  • GDANet + WOLFMix
python GDANet/main_cls.py --eval_corrupt=True --model_path pretrained_models/GDANet_WOLFMix.t7
  • RPC + WOLFMix (final)
python PCT/main.py --exp_name=test --num_points=1024 --use_sgd=True --eval_corrupt=True --model_path pretrained_models/RPC_WOLFMix_final.t7 --test_batch_size 8 --model RPC

Part Segmentation

Coming soon.