Melanoma Detection via Convolutional Neural Network trained on the HAM10000 Dataset
Access the Dataset Here: https://drive.google.com/file/d/1h91HNUN5UBE6CexoxI_eXIb-KSDZk85m/view
Before running the code, ensure paths direct to the images above.
Various Keras Models were trained on images of skin lesions from the above dataset.
The goal of the model building process was to detect the presence of melanoma from these images of skin lesions. The model detected melanoma with a precision of 0.52, recall of 0.79 and f-score of 0.62
The best results / neural network configuration can be found in the notebook "Best Results.ipynb".
To view the final results - check the Results and Charts.png file.
"Auto Cropping Lesions.ipynb" is a separate task where I attempt to crop and centre the images of skin lesions.