Multi-Class Image Classification Deep Learning Model for ISIC Challenge 2018 Using TensorFlow Take 2

Template Credit: Adapted from a template made available by Dr. Jason Brownlee of Machine Learning Mastery.

SUMMARY: This project aims to construct a predictive model using a TensorFlow convolutional neural network (CNN) and document the end-to-end steps using a template. The ISIC Challenge 2018 dataset is a multi-class classification situation where we attempt to predict one of several (more than two) possible outcomes.

INTRODUCTION: The International Skin Imaging Collaboration (ISIC) is an international effort to improve melanoma diagnosis, sponsored by the International Society for Digital Imaging of the Skin (ISDIS). The ISIC Archive contains the most extensive publicly available collection of quality-controlled dermoscopic images of skin lesions. Since 2016, ISIC has sponsored annual challenges for the computer science community associated with leading computer vision conferences.

For this modeling project, we will predict whether an image represents one of seven disease conditions (Task 3: Lesion Diagnosis). In this iteration, we will construct a CNN model based on the ResNet152V2 architecture to make predictions.

ANALYSIS: In this iteration, the ResNet152V2 model’s performance achieved a categorical accuracy score of 84.63% after ten epochs using the training dataset. The same model processed the validation dataset with a categorical accuracy rate of 83.42%.

CONCLUSION: In this iteration, the ResNet152V2-based CNN model appeared to be suitable for modeling this dataset. We should consider experimenting with TensorFlow for further modeling.

Dataset Used: ISIC 2018 Challenge Dataset

Dataset ML Model: Multi-Class classification with numerical attributes

Dataset Reference:

One potential source of performance benchmarks:

The HTML formatted report can be found here on GitHub.