Inspiration

We were inspired by the idea that information could be hidden in plain sight, without relying on conventional methods such as QR codes, barcodes or written text. We wanted to explore whether an ordinary-looking piece of abstract art could secretly carry a message that only a computer could understand.

This led us to combine artistic expression with data encoding, exploring how colours, shapes and directions could become a visual language for communicating information.

What it does

Hidden in Hue transforms a secret three-word message into an abstract piece of artwork containing a Mandela of flower blossoms.

How we built it

Message encoding: Each of the three words is mapped to an index in a dictionary of 30,000 words. These indices are combined into a single integer.

Error correction: The integer is converted into bytes and passed through a Reed–Solomon encoder, adding redundant data to help recover the message if some symbols are misread.

Visual encoding: Each byte is represented using two base-27 digits. Each digit maps to a unique combination of colour, shape and direction.

Artwork generation: Using Python and Pillow, we generate an abstract spiral containing the encoded symbols, alongside decorative organic patterns.

Decoding: The intended decoder uses computer vision to identify and classify the symbols before applying error correction and converting the recovered data back into words.

Challenges we ran into

One of our biggest challenges was balancing artistic freedom with reliable information recovery. The artwork needed to look like abstract art rather than an obvious data carrier, while keeping individual symbols distinguishable enough for a computer to recognise. Another challenge was accounting for errors. A photograph may introduce blur, poor lighting, colour variations or partially obscured symbols. Incorporating Reed–Solomon error correction helped address misclassification, although reliable photographic decoding still requires careful image processing and testing.

Accomplishments that we're proud of

We are particularly proud of developing a system that connects abstract artwork with practical data encoding.

Designed a visual alphabet containing 27 unique symbols using just three properties.

Developed a method for encoding three words into a compact numerical representation.

Incorporated Reed–Solomon error correction to improve resilience against incorrectly identified symbols.

Generated an abstract spiral artwork programmatically using Python.

Explored how organic patterns and decorative elements could be incorporated without compromising the underlying encoding structure.

Most importantly, we demonstrated how an artwork can act as a carrier for information without displaying the message itself.

What we learned

We learned that designing something for a human to appreciate and a computer to interpret presents two very different sets of challenges. A visually attractive design is not necessarily easy to decode, particularly when viewed through a phone camera under uncontrolled conditions.

What's next for Hidden in Hue

Our next steps include:

Computer vision: Build the image-processing pipeline to detect the artwork, locate the symbols and classify their colours, shapes and directions.

Robustness testing: Test under different lighting conditions, camera distances, rotations and image quality.

Improved artwork generation: Experiment with more organic spirals, flowing patterns, varied line thicknesses and richer visual compositions.

Share this project:

Updates

Submission history