
Accelerating Toward Smart 3D Scene Creation
August has been a landmark month for the SENSO3D project, as we dove head-firstinto AI model development and massive data collection. Building on the foundational work from July, our technical teams are now actively developing, testing, and refining the core machine learning pipelines that will power our AI-driven 3D object library and scene generation tools.
From building 2D datasets to training early object recognition models, this month was all about preparing our AI engine to bring immersive environments to life.
Curating the Visual Data That Fuels AI
Each image is being annotated with segmentation masks and bounding boxes to help the AI detect, classify, and eventually reconstruct the objects in 3D.
Early AI Models in Testing
We’re excited to share that early-stage AI models are now operational! These foundational models are undergoing initial testing to evaluate their performance across three key tasks:
Object Detection
Our AI can already detect and classify objects across 10 categories, with more being added every week. Using datasets like OMNI3D as inspiration,we’re customizing our models for XR applications.
2D-to-3D Shape Reconstruction
We’ve launched our first version of the reconstruction pipeline, which takes 2D images and begins generating placeholder 3D mesh structures. These will become fully textured and interactive in later stages
Metadata Generation
Beyond just creating geometry, we’re also generating metadata—such as material type,dimensions, object category, and use context—for each asset. This will make our library fully searchable and filterable.
Introducing Structured-Light Scanning
To complement our AI pipeline, we’ve begun experimenting with structured-light scanning—a high-resolution, non-intrusive technique to create hyper-accurate 3D models
We’re currently scanning selected physical objects to:
Provide ground truth 3D references for AI training
Supplement AI-generated models with scanned alternatives
Ensure high geometric fidelity in our Unity-ready scenes
This dual-track approach (AI + scanning) gives SENSO3D a powerful advantage in both automation and accuracy

Looking Ahead
September promises to be even more exciting as we:
We’ll also start planning for our first interactive scene prototype, giving users a real glimpse of what SENSO3D will offer.
Stay Tuned!
We’re on track, ahead of schedule, and already seeing the transformative potential of combining AI with immersive technology. Be sure to follow our updates for behind-thescenes insights and demos!