semantic-segmentation

Final Project for Intro To Deep Learning ECGR 5106: Use segmentation to determine and locate tumor in MRI scan

Path for training a semantic segmentation model with encoder-decoder:

1. Test CNN Encoder-Decoder
2. Add skip connections
3. Add Regularization: -BatchNorm, Dropout, L1, L2

Image Source: https://www.google.com/url?sa=i&url=https%3A%2F%2Fcnvrg.io%2Fsemantic-segmentation%2F&psig=AOvVaw3C6hrrIJRAKAsP_RVwhUMH&ust=1712790866826000&source=images&cd=vfe&opi=89978449&ved=0CBIQjRxqFwoTCLD4jq2htoUDFQAAAAAdAAAAABAE

Classification vs Segmentation

Image Source: https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2021.638182/full

Encoder-Decoder Network Source: https://towardsdatascience.com/understanding-u-net-61276b10f360

ConvNeXt Source: https://github.com/yassouali/pytorch-segmentation/blob/8b8e3ee20a3aa733cb19fc158ad5d7773ed6da7f/models/segnet.py#L9

Pre-Processing Data for CNN Source: https://towardsdatascience.com/how-to-apply-a-cnn-from-pytorch-to-your-images-18515416bba1

Dataset Source: https://www.kaggle.com/datasets/pkdarabi/brain-tumor-image-dataset-semantic-segmentation/code