[AR]t

Inspiration:

We were inspired by the inconvenience faced by novice artists creating large murals, who struggle to use reference images for guiding their work. It can also help give confidence to young artists who need a confidence boost and are looking for a simple way to replicate references.

What it does

An AR and CV based artist's aid that enables easy image tracing and color blocking guides (almost like "paint-by-numbers"!) It achieves this by allowing the user to upload an image of their choosing, which is then processed into its traceable outlines and dominant colors. These images are then displayed in the real world on a surface of the artist's choosing, such as paper or a wall.

How we built it

The base for the image processing functionality (edge-detection and color blocking) were Python, OpenCV, numpy and the K-means grouping algorithm. The image processing module was hosted on Firebase. The end-user experience was driven using Unity. The user uploads an image to the app. The image is ported to Firebase, which then returns the generated images. We used the Unity engine along with ARCore to implement surface detection and virtually position the images in the real world. The UI was also designed through packages from Unity.

Challenges we ran into

Our biggest challenge was the experience level of our team with the tech stack we chose to use. Since we were all new to Unity, we faced several bugs along the way and had to slowly learn our way through the project.

Accomplishments that we're proud of

We are very excited to have demonstrated the accumulation of our image processing knowledge and to make contributions to Git.

What we learned

We learned that our aptitude lies lower level, in robust languages like C++, as opposed to using pre-built systems to assist development, such as Unity. In the future, we may find easier success building projects to refine our current tech stacks as opposed to expanding them.