Brain-Tumor-Segmentation

Semantic Segmentation of tumor from Brain MRI images using PyTorch.

The model architecture used is UNET which is trained using PyTorch, and then converted to ONNX format for deployment using Heroku. Evaluation metric used is DICE coefficient, with loss as (1-DICE) + BCELoss.

Dataset 📂

Dataset used for training is from Kaggle LGG Segmentation Dataset which which contains over 3900 samples obtained from The Cancer Imaging Archive.

Notebook 📒

View the notebook here: brain_tumor_segmentation.ipynb

Deployment 🚀

The model has been converted to ONNX format and deployed using Gradio & hosted on Heroku: Brain MRI Tumor Detection

Predictions 🔍

Predictions on unseen test data: