<aside> đź“–

The final project should be a creative project that builds off of or is inspired by the concepts we've covered this semester. You should feel free to think non-traditionally, projects do not need to be screen-based and there is no requirement to use a particular programming language or machine learning model.

Other notes:

  1. Keeping things simple and small in scope is a plus. If your project idea is a big one, consider documenting the larger idea but implementing just a small piece of it.
  2. Also think about making a final project for a small audience, even one single person like a family member or friend. . . or yourself. This can be a good way to focus your idea and design process. "Generalizing" the idea can come later (or maybe not at all.) </aside>

Project Introduction

Project title: 3Depth

One sentence summary: Creating a system, which allows the user to view the depth data of an uploaded image in a 3D canvas.

Project Abstract

I would like to create a system, which would allow its users to view the depth estimation data of an image in an interactive 3D space, which can be rotated freely using a mouse.

I would like it to have an intro screen where you can drag and drop an image onto, and then it would it would analyse the image, and display it in a grid system.

I am planning to utilise the HuggingFace based depth-estimation example in p5, and create a 3D canvas to display the data.

I would like to start small, and get it working with static images, but later on I think it could be an interesting challenge to see, whether videos could be something the system can process, as they are “just a sequence of images” in theory.

FUTURE SUGGESTIONS: Additionally, I wouldn’t assume I will have time for it unfortunately in this project, but I am thinking of different ways to utilise this point-cloud like structure. I think it could be really interesting for example to make the dots vibrate more or less based on the depth from the captured camera. Or maybe using multiple cameras from different angles could create a coherent point cloud which truly would be a 360 representation / scan of the object. Irregardless, I believe this project could be a great test-bed / foundation to build on top of in the future.

Inspiration Material

One of my previous code-explorations included working with the depth estimation example. I found if fascinating, and I wanted to continue working with it in this project as well.

Also, in the past I have found a lot of people working with point clouds in various different 3D software, making amazing art with them, which always looks really cool.