AtomicXR: Capture atomic structures to build 3D in MR.

5 minutes

Full Video

Introducing AtomicXR

AtomicXR is a Mixed Reality project that combines the Meta Passthrough Camera API with the power of OpenCV for real-time image processing. The concept is simple yet impactful: Pinch → Capture → Detect → Generate.

Using a hand gesture (like a pinch and drag), users can capture a sketch or printed image of an atomic structure. The system then processes the image to detect and analyze the number of electrons or elements using computer vision techniques. Based on this detection, a 3D atomic model is generated and visualized right in your physical space through Mixed Reality.

This project blends spatial interaction with educational visualization — making complex atomic concepts interactive and easier to understand in real time.

Key Features

Natural World Capture: A smooth and intuitive interaction system allows users to pinch and drag using hand gestures to capture atomic structure sketches from the real world using the Passthrough camera. This makes the interaction feel natural, creative, and human-centered.
Advanced Detection Algorithm: The app leverages multiple layers of OpenCV image processing to accurately analyze atomic structure drawings. Through several iterations, it extracts key visual cues — such as the number of electrons — and determines which element the user has drawn.
Dynamic 3D Structure Generation: Based on the detected electron count, the system generates a 3D atomic model using rings and spheres. Each component is placed and angled correctly to represent the element's atomic structure in 3D space, offering a clear and interactive visualization.

Step 1: Pinch to capture a hand-drawn atomic structure using the Passthrough camera.

Step 2: Image is processed to detect electrons, then dropped onto a surface like table or floor.

Step 3: A 3D atomic model is generated based on the detected structure.

Behind the AtomicXR Development

The concept for this project began back in 2021 as part of my final year college project. At that time, I was experimenting with Augmented Reality on Android phones — capturing atomic structure images and analyzing them with OpenCV to generate basic 3D models. (Link to old project)

Later, when the Meta Passthrough Camera API was released, the idea sparked again. I realized this could be the perfect opportunity to bring my old project into a more immersive Mixed Reality environment.

That’s how AtomicXR was borns — a fresh take on the original "Atomic" project, now rebuilt with hand interactions, spatial understanding, and real-time 3D generation in MR.

Step-by-step breakdown of how OpenCV processes captured images in Unity to detect and count electrons in atomic structures.

Watch the system auto-generate 3D atomic models based on the detected number of electrons.

Demonstrates how intuitive pinch and drag gestures create a real-time spatial frame, defining the capture area for atomic sketches.

Using the frame gesture, a mesh is created with grab scripts, allowing users to intuitively interact with their selected area.

Future Development

While the core concept of AtomicXR works well, I’ve identified a key limitation — relying solely on OpenCV makes the detection process inconsistent, especially under varying lighting conditions or with different drawing styles and image types.

To make the system more scalable and robust, I plan to move from traditional image processing to a custom-trained AI model. I'm also exploring the use of Gemini or ChatGPT APIs for more intelligent analysis. This shift will not only improve accuracy but could also provide real-time feedback or suggestions to correct the drawing — adding a valuable learning layer to the experience.

Aman Bohra

Hi there! I'm Aman Bohra, the developer behind AtomicXR. I'm passionate about creating innovative mixed reality experiences that push the boundaries of what's possible in interactive technology.