Inspiration
Millions of people live with Color Vision Deficiency (CVD), struggling to interpret colors accurately in everyday life. We were inspired to use AI and deep learning to bridge this gap, creating a solution that restores natural color perception and visual confidence for colorblind users.
What it does
AutoLens is an AI-driven assistive tool that corrects colorblind images in real-time. Users can upload images through a web interface, and our system outputs color-corrected visuals while preserving texture and structural details, improving accessibility and inclusion.
How we built it
1) Model Training
- Trained a custom U-Net model using TensorFlow/Keras on Color Blindness Simulation & Correction dataset
- Preprocessed and augmented data for robust learning.
- Achieved a model capable of transforming colorblind images to near-natural vision.
2) Web Deployment (app.py)
Deployed the trained model via Flask for real-time image correction.
Features image upload, preprocessing, inference, and output comparison.
Displays metrics like PSNR (Peak Signal-to-Noise Ratio) and SSIM (Structural Similarity Index) for evaluation.
Together, these components form a complete end-to-end AI pipeline, from training to live deployment.
Challenges we ran into
Balancing color restoration with texture preservation.
Optimizing model size and inference speed for smooth web performance.
Fine-tuning data preprocessing and loss functions for realistic results.
Accomplishments that we're proud of
Built a robust AI model capable of transforming colorblind vision into a near-natural experience.
Successfully deployed the model on a user-friendly web interface for real-time interaction.
Demonstrated how AI + Cloud + Assistive Technology can directly impact accessibility.
What we learned
Designing U-Net architectures for image-to-image tasks.
Integrating deep learning with web deployment using Flask.
Applying SSIM and PSNR metrics for visual quality assessment.
Managing datasets and pipelines for cloud-ready AI applications.
What's next for AutoLens: AI-Powered Vision
Expand support for different types of colorblindness.
Improve mobile accessibility with app integration.
Optimize the model for faster inference and higher accuracy.
Explore real-time video correction for enhanced user experience.
Log in or sign up for Devpost to join the conversation.