EmoCanvas

Inspiration Art has always been a reflection of human emotion, but not everyone can translate feelings into visuals. We wanted to bridge the gap between raw emotion and creative expression using AI. Inspired by tools like DALL-E and advancements in facial recognition, EmoCanvas detects emotions through a webcam and generates unique art in real-time, empowering users to see their moods come alive on screen.

What We Learned

Real-time emotion detection using pre-trained models (FER-2013 dataset).

Fine-tuning generative AI models (Stable Diffusion) for stylistic consistency.

Optimizing latency for seamless user experience.

How We Built It

Emotion Detection: A lightweight CNN model processes live video to classify emotions (happy, sad, angry, etc.).

AI Art Generation: Stable Diffusion is conditioned on the detected emotion to create abstract or realistic art.

Interactive UI: A React frontend lets users tweak styles (e.g., "watercolor" or "cyberpunk") and save/share their art.

Challenges

Balancing model accuracy with real-time performance.

Ensuring diverse art outputs without repetitive patterns.

Handling edge cases (e.g., low-light environments).

Built With

• cloudinary
• fastapi
• fastapi-ai-models:-tensorflow-(emotion-detection)
• firebase-(user-storage)-apis:-webcam-capture
• javascript
• javascript-frameworks:-react
• python
• react
• stable-diffusion-(art-generation)-cloud:-aws-ec2-(hosting)
• stablediffusion
• tensorflow