Skin Lesion Analysis for Melanoma Detection
Set of images of skin lesions intended for the training of algorithms for the automatic detection of melanoma, with real cases.
Description
This dataset contains 2750 high-resolution images of various skin lesions, collected to aid in the automated detection of melanoma. It was used as part of an international challenge (ISIC 2017) and is a reference in the field of AI-assisted dermatological imaging.
What is this dataset for?
- Train medical image classification models (benign vs melanoma)
- Testing algorithms for the detection of skin pathologies based on clinical photos
- Create AI solutions for teleconsultation or preliminary diagnosis in dermatology
Can it be enriched or improved?
Yes, it is possible to complete this dataset with patient metadata (age, gender, history), or to annotate specific suspicious areas using annotation tools. Data augmentation techniques can also improve performance in rare cases.
🔎 In summary
🧠 Recommended for
- Researchers in assisted dermatology
- AI health startups
- Computer vision students
🔧 Compatible tools
- Keras
- PyTorch
- TensorFlow
- MONAI
- Label Studio
💡 Tip
Balance the benign/melanoma classes via oversampling or targeted augmentation for better results.
Frequently Asked Questions
Does this dataset contain medical diagnoses associated with each image?
No, only images are included. It is recommended to cross-reference them with clinical metadata for advanced use.
Is this dataset suitable for mobile detection projects?
Yes, with an optimized model, this dataset can be used to prototype a mobile diagnostic or teleconsultation app.
Is this data set sufficient for a complete workout?
It can be used as a base, but a robust model will benefit from a complement with other ISIC datasets or data augmentation.
