ResearchXR2024 — 2025

iXR: Intelligent Spatial Computing

An EU-funded (Horizon Europe) spatial-computing and XR platform that reads the real world, built with EMIL and Aalto University.

EMIL project page

iXR ("Intelligent Spatial Computing and XR development platform") is Reality Crisis's project in EMIL, the European Media and Immersion Lab. EMIL is funded by the EU's Horizon Europe programme (grant agreement 101070533), and Aalto University is one of its partners.

The goal is to grow Skatrix's real-world sensing into a platform that other developers can build XR experiences on. It combines AR, AI and geolocation, so virtual content lays itself over real geometry in a believable way, well beyond skateboarding.

My role

LeadArchitect of the object-detection and scene-analysis system

Team

Reality Crisis · 2019 — present

Platforms

iOS
Android
XR

Stack

Unity 2023.2
C#
AR Foundation
ARCore
Unity Sentis
YOLOv8 (ONNX)
Ultralytics · Roboflow
Niantic Lightship VPS
OpenCV

Lead

What I did

01Architected the YOLO-driven object detection, running on-device through ONNX Runtime.
02Built the scene analysis that measures ledges, stairs, grind surfaces and gap distances.
03Connected detection to gameplay: trick suggestions, approach speed and entry angles per segment.
04Delivered the project inside EMIL, the EU Horizon Europe programme with Aalto University.

What stands out

01The system measures the real world: heights, widths and lengths of ledges, stairs and grind surfaces, with the distances marked for gap jumps.
02It plans tricks from what it sees, suggesting trick variations, approach speed and entry angle for each segment, and finding skate lines from the surrounding geometry.
03An AR skater then rides the planned line, with info tags beside each trick and an interactive path highlight.
04YOLO object detection on-device through ONNX Runtime, feeding real-time scanning and context-aware interactions.
05Additional field-test footage: AI trick-suggestion logging via a Hugging Face API call surfaced in the in-app debug console, a Monk/Place-Skater trick-selection screen over a scanned mesh spot, VPS spot spawn-point selection, an AR skater placed live in a room answered by the 'Monk' AI assistant, and a separate 2D-image-to-animated-3D-character-to-AR placement demo (CSM AI-style pipeline).