Data

We trained the program with online Dermoscopic images of melanoma that were zoomed in, but very high quality.

Access our Dataset

Link to our Dataset: Press Here
Additional Datasets we Analyzed:
- ISIC Archive: Press Here
- VisualDX: Press Here

We wanted the main focus of the image to be the unhealthy skin so that the program would have a very good idea of what melanoma looks like, and not take the surrounding skin into account as much when identifying it.

Image Courtesy of: https://en.wikipedia.org/wiki/Melanoma

However, melanoma is a disease that can look very different from case to case, so we focused our algorithm on detecting distinctly melanated melanoma with varying degrees of unevenness in dermoscopic images.

Image Courtesy of: https://www.healthline.com/health/melanoma-pictures

When focusing on the melanoma in the image, we included melanomas of different shape, size, coloring, texture, and surrounding skin color. However, we also incorporated images that have slighting different lighting conditions and angles.

Image Courtesy of: https://www.healthline.com/health/foot-melanoma