Description :

this deep learning model was made to classify skin face disease including acne, wrinkle, and redness using tensorflow we build CNN model that also implement transfer learning and fine tuning

Packages and Library :

• os
• numpy
• pandas
• seaborn
• matplotlib.pyplot
• tensorflow

Dataset We Used :

https://www.kaggle.com/datasets/panupongsingdee/acneclearcomedo-dataset

Model Architecture :

What We Used In Our Model:

1. Transfer Learning:

   • Menggunakan MobileNetV2 sebagai backbone model.

2. Pooling:

   • GlobalAveragePooling2D digunakan untuk merangkum fitur dari MobileNetV2 menjadi vektor.

3. Fully Connected Layer:

   • Dense layer dengan 128 neuron menggunakan fungsi aktivasi ReLU.

4. Dropout:

   • Dropout sebesar 0.5 digunakan untuk mencegah overfitting.

5. Output Layer:

   • Dense layer dengan 4 neuron menggunakan fungsi aktivasi Softmax untuk prediksi kelas.

Training Result :

Training Parameter :

• Optimizer: Adam
• Learning Rate: 0.0001
• Loss Function: Categorical Crossentropy
• Batch Size: 32
• Epochs: 50
• Validation Split: 20%

Accuracy :

• Training Accuracy: 0.9156%
• Validation Accuracy: 0.9060%
• Training Loss: 0.2412%
• Validation Loss: 0.2676%

Test Evaluate :

• Test accuracy : 0.8833%
• Test loss : 0.3522%

Confusion Matrix :