Many of us struggled with acne growing up, navigating a sea of online misinformation and influencer-driven marketing. Products were promoted without transparency, while effective, affordable options were often overlooked. This impacted not just our skin, but our self-confidence and engagement with the world.
EasyDerma challenges the idea that acne is just a natural, unavoidable phase—and aims to give people clarity, confidence, and control.
Users upload a facial image, and EasyDerma's custom-trained AI model diagnoses visible skin conditions. A second opinion is generated via the Gemini API to confirm the analysis. Based on this consensus, EasyDerma generates:
- A personalized skincare routine (AM & PM)
- Product recommendations with Amazon links
- Usage instructions and explanations
- Dietary tips to support skin health holistically
- Frontend: Next.js with React and TypeScript
- Authentication: Auth0 for secure user login
- Backend: Flask connected to MongoDB
- ML Model: Custom CNN trained on a Kaggle dataset of 27,000+ dermatological images
- AI Services: Gemini API integration for second-opinion diagnostics
The CNN includes 6 convolutional layers with max pooling and data augmentation to improve accuracy and reduce overfitting using TensorFlow and Keras.
- Managing dataset inconsistencies
- Hardware constraints for model training
- Integrating AI results with dynamic UI
- Storing and retrieving image data effectively
- Built and trained a custom CNN from scratch
- Created a synced, cross-platform experience using MongoDB
- Integrated live diagnosis + product suggestions in a clean and interactive UI
- The practical limitations of applied machine learning
- Hands-on experience with Next.js and full-stack integration
- Implementation of Auth0 for secure flows
nextjsreacttypescriptjavascriptflaskmongodbauth0geminitensorflow / keras
- GitHub: @SecretariatV
- Email: [email protected]
- Telegram: @ares_orb
- Twitter (X): @OVB_Coder