README.md

Medical Image Segmentation

Team: Autobots

Team Members:
Abhinaba Bala (2020701001)
Neelabh Kumar (2020701003)
Ruchi Chauhan (2018711001)
Rupak Lazarus (2020701020)

Assigned TA: Sai Soorya Rao Veeravalli

MidEval Presentation File

Final Presentation File

This project is undertaken as a part of the Computer Vision coursework at IIIT Hyderabad in Spring semester 2021. The main inspiration of our work is from the very recent paper of TransUNet which merits both Transformers and U-Net, as a strong alternative for medical image segmentation. The overview and proposal can be found here.

Code structure:

-lists
-networks
CVmidEval.pdf
ProjectProposal.pdf
README.md
ScatterringCoeff_generator.m
US_generateMasks.py
Unet.ipynb
eval.py
rc_bcet.py
train.py
utils.py

Dataset

Download the dataset,
For lung images: here
For ultrasound images: here

Pre-processing

To generate the scattering-coefficients for augmenting in TranUnet, use ScatteringCoeff_generator.m
For processing the ultrasound images use the US_generateMasks.py file.
And, for processing the lung images use rc_bcet.py file.

TransUNet Eval

python3 eval.py --dataset Ultrasound --batch_size 12 --model_path /ssd_scratch/cvit/rupraze/models/ultrasound/epoch_120.pth

TransUNet + Scattering coefficients Eval

python3 eval_scatcoeff.py.py --dataset Ultrasound --batch_size 12 --model_path /ssd_scratch/cvit/rupraze/models/scat_ultrasound/epoch_149.pth

Inference

To get the demo of the inference process, run the 'InferenceDemo.ipynb' file