Melanoma Skin Cancer diagnosis using Image Classification of skin images using Deep Learning and Image Processing. This repository uses Deep Learning CNN Model built on PyTorch and trained over Google Cloud TPU.
Currently working on optimizing the results for Competition Submission. Will update the final version soon. Stay Tuned😀 .
While skin cancer is the most prevalent type of cancer, Melanoma, specifically, is responsible for over 75% of skin cancer deaths, despite being least common skin cancer. Dermatologist currently have to diagnose the patients by evaluating patients every mole to find the melanoma skin lesions.
Dermatologists can accurately diagnose Melanoma with the help of Artificial Intelligence and Data Science. AI can offer a higher amounts of accuracy and provides an extra validation for human-based diagnosis.
- Melanoma Classification is done using the data provided by the joint effort of Society for Imaging Informatics in Medicine (SIIM) and International Skin Imaging Collaboration (ISIC) in their Kaggle Competition.
- Melanoma Classification Dataset - Preprocessed with external Dataset by Roman
- PyTorch
- PyTorch XLA for TPU Support
- TorchVision
- EfficientNet by Google AI (Deep Learning CNN Architecture)
- Albumentation
- WTFML
- Google Cloud TPU
- Google Colab
- Sincere thanks to Chris Deotte and Abishek Thakkur for their TPU Power Hour. It helped me to learn training Deep Learning Model over TPUs.