---
license: openrail
---

***Disclaimer: This work is part of an educational project. It is not intended for clinical application. As such it can not make real world predictions for skin lesions. To get recommendations regarding skin lesions one should ask for expert advice such as provided by a dermatologist.***

The model (xception_v4_1_07_0.699.h5) was trained as described in this kaggle notebook: https://www.kaggle.com/bnzn261029/capstone1-ham10k-skincancer

The code repository on github: https://github.com/bsenst/capstone1-skin-lesion-classifier

The dataset on kaggle: https://www.kaggle.com/datasets/kmader/skin-cancer-mnist-ham10000

The gradio app on huggingface spaces: https://huggingface.co/spaces/bsenst/keras-image-classifier

|Layer (type)|Output Shape|Param|  
|-|-|-|
|input_2 (InputLayer)|[(None, 150, 150, 3)]|0|
|xception (Functional)|(None, 5, 5, 2048)|20861480|
|global_average_pooling2d (GlobalAveragePooling2D) |(None, 2048)|0|             
|dense (Dense)|(None, 7)|14343|
Total params: 20,875,823
Trainable params: 14,343
Non-trainable params: 20,861,480