Melanoma Skin Cancer Detection using Convolutional Neural Networks and Transfer Learning

This is a Kaggle competition in which we have to identify if the given lesion image is malignant or not for Melanoma which is a type of skin cancer.

📝 Table of Contents

🧐 About
📊 Dataset Overview
🧠 Model Building
🎯 Model Performance
🏅 Model Evaluation

🧐 About

Melanoma is a type of skin cancer that occurs when pigment-producing cells (melanocytes) mutate and become cancerous. Melanoma is the most serious form of skin cancer and 5th most common cancer.

The American Cancer Society estimates around 207,390 new melanoma cases will be diagnosed in 2021. As with other cancers, early and accurate detection—potentially aided by data science—can make treatment more effective.

Currently, dermatologists evaluate every one of a patient's moles to identify outlier lesions that are most likely to be melanoma.

📊 Dataset Overview

The dataset was taken from the competition hosted by SIIM-ISIC on Kaggle.

Malignant: A cancerous growth that can spread very quickly and invade nearby tissues
Benign: A non-cancerous growth that usually grows very slowly and does not spread to other areas

Click to view 👇 :

🧠 Model Building

Here, the bottom layers of pre-trained Xception model were used and the top layers were fine-tuned for our particular task to classify skin cancer images as malignant or not. The architecture of xception model is as follows:

🎯 Model Performance

Since the dataset is highly imbalanced, accuracy wouldn't be a good metric to judge the model performnace hence we have used the AOC as a performance metric along with BinaryCrossentropy as the loss function. Here is the model's performance.

🏅 Model Evaluation

We evaluated our model on the test dataset and we were able to achieve an AUC of 0.8459. Here is the curve:

Melanoma Skin Cancer Detection using Convolutional Neural Networks and Transfer Learning🕵🏻‍♂️

Related tags

Overview

Melanoma Skin Cancer Detection using Convolutional Neural Networks and Transfer Learning

📝 Table of Contents

🧐 About

📊 Dataset Overview

Click to view 👇 :

🧠 Model Building

🎯 Model Performance

🏅 Model Evaluation

Click to view 👇 :

Thank you!

Owner

Vipul Shinde

Hl classification bc - A Network-Based High-Level Data Classification Algorithm Using Betweenness Centrality

A library for uncertainty representation and training in neural networks.

Model Zoo for AI Model Efficiency Toolkit

Deep learning image registration library for PyTorch

A Factor Model for Persistence in Investment Manager Performance

IOT: Instance-wise Layer Reordering for Transformer Structures

A python module for scientific analysis of 3D objects based on VTK and Numpy

Code of the paper "Part Detector Discovery in Deep Convolutional Neural Networks" by Marcel Simon, Erik Rodner and Joachim Denzler

This repository contains the code to replicate the analysis from the paper "Moving On - Investigating Inventors' Ethnic Origins Using Supervised Learning"

Code for our paper "Multi-scale Guided Attention for Medical Image Segmentation"

KSAI Lite is a deep learning inference framework of kingsoft, based on tensorflow lite

Code for Iso-Points: Optimizing Neural Implicit Surfaces with Hybrid Representations

FEMDA: Robust classification with Flexible Discriminant Analysis in heterogeneous data

A crash course in six episodes for software developers who want to become machine learning practitioners.

This repository contains code to run experiments in the paper "Signal Strength and Noise Drive Feature Preference in CNN Image Classifiers."

ML-Decoder: Scalable and Versatile Classification Head

Code for "LoFTR: Detector-Free Local Feature Matching with Transformers", CVPR 2021

Implementation of Memory-Efficient Neural Networks with Multi-Level Generation, ICCV 2021

TransFGU: A Top-down Approach to Fine-Grained Unsupervised Semantic Segmentation

Guiding evolutionary strategies by (inaccurate) differentiable robot simulators @ NeurIPS, 4th Robot Learning Workshop