Photofeeler

An innovative Deep Learning system for automated assessment of perceived likability of people in photographs. This project combines Machine Learning and Deep Learning techniques to master the complex task of sympathy recognition. As part of an extensive study with 188 participants, 350 images were evaluated for sympathy, beauty, image quality, and emotions. Based on this dataset, a Machine Learning model was developed that automatically extracts 61 relevant image features – including facial emotions, attractiveness, body posture, and image quality. The final Random Forest model achieves impressive accuracy and can reliably distinguish between sympathetic and unsympathetic images. The application was implemented as a complete full-stack solution with React frontend, FastAPI backend, and client-side end-to-end encryption. Users can securely upload their images, have them analyzed, and receive detailed recommendations for improving their photo impact.

• AI-powered sympathy recognition
• Automatic extraction of 61 image features
• Facial emotion recognition
• Attractiveness analysis
• End-to-end encryption (AES-256-GCM)
• Personalized improvement recommendations