Run the script:
cd ./gapro
python3 gen_ps.py --save_folder dataset/scannetv2/gaussian_process_kl_pseudo_labels
The pseudo labels will be stored in dataset/scannetv2/gaussian_process_kl_pseudo_labels
1) First navigating to ISBNet subfolder:
cd ISBNet/
2) Pretrain the 3D Unet backbone from scratch
python3 tools/train.py configs/scannetv2/boxsup_isbnet_backbone_scannetv2.yaml --only_backbone --exp_name pretrain_backbone
3) Train ISBNet
python3 tools/train.py configs/scannetv2/boxsup_isbnet_scannetv2.yaml --trainall --exp_name default
4) (Optional) Self-training
Generate pointwise deep features from the current best model
python3 tools/export_features.py configs/scannetv2/boxsup_isbnet_scannetv2_export_feats.yaml <checkpoint_file> --save_deepfeatures_path dataset/scannetv2/pretrain_maskfeats
Re-generate the pseudo labels:
cd ./gapro
python3 gen_ps.py --save_folder dataset/scannetv2/gaussian_process_kl_deep_pseudo_labels --use_deepfeat --deepfeat_folder dataset/scannetv2/pretrain_maskfeats/
Change the label_type in config file to the new label name (gaussian_process_kl_deep_pseudo_labels) and re-run step #2 and #3
1) For evaluation (on ScanNetV2 val, S3DIS, and STPLS3D)
python3 tools/test.py configs/<config_file> <checkpoint_file>
2) For exporting predictions (i.e., to submit results to ScanNetV2 hidden benchmark)
python3 tools/test.py configs/<config_file> <checkpoint_file> --out <output_dir>