All packages and dependencies are included in the requirements.txt.

• 3.7

Without a doubt, most of teenages face acne problem. However, there isn't a guideline on how serious the acne is and what proper steps should be taken in order to cure the acne and prevent scars. Hence, acne classification is developed using deep learning. It is carried out using Resnet-18 model in this project. It is able to classify acne seriousness into:

1. Normal
2. Level 0
3. Level 1
4. Level 2

as shown below:

250 HD images were being hand-picked for each classes from various internet sources.

Data annotation is being carried out by separating dataset into 4 classes.

To increase the size of the dataset for training, data preprocessing is being carried out which includes:

• Flip: Horizontal, Vertical
• 90° Rotate: Clockwise, Counter-Clockwise, Upside Down
• Crop: 0% Minimum Zoom, 50% Maximum Zoom
• Rotation: Between -15° and +15°
• Blur: Up to 10px
• Rotate: 30 degree

pip install -r requirements.txt

streamlit run main.py

The AI Model is able to achieve up to 90% accuracy by training with only 250 HD images from each classes.

Accuracy	Loss Function	Confusion Matrix

1. Platform to discuss skin care products
2. Cross geographical skin samples
3. Develop smartphone app
4. More detailed classifier
5. Higher Accuracy
6. Implement Object Detection
7. Suggest Possible Treatments
8. Consult Dermatologist Virtually

Mobile App	Web App

Additional information about this project can be read here.