Skin Cancer Detection: Deep-Learning Melanoma Detection

We developed computer-vision classification workflows using CNN and ensemble approaches and transfer learning, combined with demographic information. Pretrained architectures such as ResNet and EfficientNet were part of the toolset, with OpenCV for image processing.

• ProblemImages needed consistent preprocessing.
  →
  Our answerWe built preprocessing into the workflow with OpenCV.
• ProblemClassification needed strong models without training from scratch.
  →
  Our answerWe used transfer learning with CNN and ensemble approaches.
• ProblemImages alone may not tell the whole story.
  →
  Our answerWe combined demographic information with image classification.

The project delivered an automated research workflow for melanoma classification.

• Automated melanoma image classification research workflow
• CNN, ensemble and transfer-learning models in one pipeline
• Demographic information included in classification