Web-based Melanoma Detection

ArXiv link

Standalone Melanoma web application

Melanoma web application

Supported datasets

• HAM10000
• ISIC2016
• ISIC2017
• ISIC2018
• ISIC2019
• ISIC2020
• PH2
• 7-point criteria dataset
• PAD_UFES_20
• MED-NODE
• Kaggle

SIIM-ISIC leaderboard results

DBs	Network	Private Score 1	Public Score 2
ISIC16'+ISIC17'+ISIC18'+ISIC19'+MEDNODE+Kaggle	DenseNet121	0.7211	0.7472
ISIC18'	ResNet50	0.5999	0.6301
ISIC20'	ResNet50	0.7751	0.8126
ISIC16'+ISIC17'+ISIC18'+ISIC19'+ISIC20'+PH2	ResNet152	0.8064	0.8073
ISIC19'	ResNet152	0.6769	0.7234
ISIC16'+ISIC17'+ISIC18'+ISIC19'+ISIC20'+PH2+ 7pointcriteria+PAD_UFES_20+MEDNODE+Kaggle	ResNet152	0.7774	0.7894
Multiple 3	Ensemble 4	0.7618	0.7621

Note

¹ Score on 70% of private testsets. The potential winner(s) are determined solely by the leaderboard ranking on the private leaderboard.
² Score on the public testsets for reference
³ Averaged the models in the table, trained with multiple datasets
⁴ Averaged the probabilities from the models in the table

Environment (chimera clean env)

• Keras - 2.5.0rc0
• Tensorflow - 2.5.0
• Augmentor - 0.2.10
• matplotlib==3.2.1
• pandas==1.2.0
• numpy==1.19.4
• pip install pydot
• conda install -c anaconda graphviz

Contact

[email protected]