/ActiveST

Active Self-Training for Weakly Supervised 3D Scene Semantic Segmentation

Primary LanguagePython

Train model on ScanNet-v2

# 1. train a model using random sample annotations
python train_init.py

# 2. run ActiveST framework
sh train.sh

Make submission to ScanNet-v2 benchmark

# 1. make predictions
python test.py

# 2. remap label idx
python make_submit.py