AI to Enhance Art Perception for Protanopia & Deuteranopia

Inspiration

Art is a universal experience, and everyone deserves the opportunity to engage with it fully. Yet, for a significant portion of the population, those with color blindness, experiencing art can be limited, especially when it comes to red and green hues. These two colors constitute approximately 35–40% of artworks, respectively, making them crucial elements in visual art. Unfortunately, this means that many paintings, illustrations, and digital pieces are often inaccessible to individuals with red-green color blindness.

What it does

We are developing an AI that dynamically adjusts and enhances the perception of red and green tones based on a novel approach inspired through research, generating pictures the colorblind are able to understand. This ensures that art becomes more inclusive and perceptually enriching for all.

How I built it

Scikit-learn provides classic machine learning models like KMeans, which are useful for color clustering and segmentation.

OpenCV (cv2) is a powerful computer vision library often used for image manipulation and filtering.

NumPy supports mathematical and matrix operations, making it ideal for processing image data.

Pillow (PIL) helps with loading and saving images, as well as converting formats and enhancing image quality.

For more advanced tasks, TensorFlow and PyTorch offer deep learning frameworks that can be used to create models for color correction.

Finally, Matplotlib and Seaborn are visualization tools useful for plotting color maps and histograms, helping to understand and visualize image data better.

Challenges I ran into

The execution of the algorithm we decided to utilize Creating a basic and accurate sample for our app Finding novel methods not yet applied in AI for color vision deficiency Balancing color accuracy while preserving artistic intent

Accomplishments that I'm proud of

Creating an algorithm based on novel research:

• That recolors images naturally
• Allows colorblind people to easily distinguish between indistinguishable colors
• Built a user-friendly interface Testing our method on red and green colorblind people and having positive and successful results.

What I learned

Colorblind users will get to have the chance to visualize or experience the difference between colors.

Empathy and awareness for non-color-blind users

What's next for AI to Enhance Art Perception for Protanopia & Deuteranopia

Implementing an evaluation test for the user to ensure a more accurate output for the audience by: Allowing users to draw the confusion line in a CIE 1931 chart, color diagram, to identify the shades of colors that they see as one, to make it more personalized and suitable for the user . Replacing the K-means clustering by fuzzy clustering to have a more precise result. Conducting further assessments of colorblind people

Built With

• as-well-as-converting-formats-and-enhancing-image-quality.-for-more-advanced-tasks
• cv2
• kmean
• making-it-ideal-for-processing-image-data.-pillow-(pil)-helps-with-loading-and-saving-images
• matplotlib
• matplotlib-and-seaborn-are-visualization-tools-useful-for-plotting-color-maps-and-histograms
• numpy
• opencv
• pil
• pytorch
• scikit-learn
• seaborn
• tensorflow
• tensorflow-and-pytorch-offer-deep-learning-frameworks-that-can-be-used-to-create-models-for-color-correction.-finally
• which-are-useful-for-color-clustering-and-segmentation.-opencv-(cv2)-is-a-powerful-computer-vision-library-often-used-for-image-manipulation-and-filtering.-numpy-supports-mathematical-and-matrix-operations