Few-shot Segmentation Experiments

Adopted from ASGNet.

---

Requisites

• Test Env: Python 3.9.7 (Singularity)
• Packages:
  • torch (1.10.2+cu113), torchvision (0.11.3+cu113)
  • numpy, scipy, pandas, tensorboardX
  • cv2

---

Clone codebase

cd /scratch/$USER
git clone https://github.com/TeamOfProfGuo/FSS-Exp
cd FSS-Exp

---

Preparation

Pascal-5i dataset

Note: Make sure the path in prepare_dataset.sh works for you.

cd /scratch/$USER/FSS-Exp
bash prepare_dataset.sh

Pretrained models

Download via this link, and transfer the zip file to your project root on Greene.

cd /scratch/$USER/FSS-Exp
unzip initmodel.zip

Config file

Modify the data_root under config/pascal/pascal_asgnet_split0_resnet50.yaml.

---

Training

Note: Modify the path in slurm scripts (as needed) before you start.

# switch to project root
cd /scratch/$USER/FSS-Exp

# train & save & test
sbatch train.slurm pascal asgnet split0_resnet50

# After the job ends:
# [To be updated]