CortexiCAM / Myriad-X
A camera-side prototype where the constraint was immediate: limited compute, real lighting, latency, and a small device that still had to make perception visible.
AI cameras
AI cameras and applied vision.
After robotics, the same discipline moved into cameras: edge devices, awkward lighting, latency, real rooms, and the need to make perception useful outside the lab.
The visible objects are CortexiCAM / Myriad-X, the UP Board edge-computing article, camera-year films, Generator, and the six-view fused-cloud precursor.
Watch demo
Camera work
Three objects from the camera years.
Each object keeps the line from robots to cameras to coordinates physical.
Room-state films
Entrance and monitoring films turned camera output into states a person could inspect in the room, not just scores in a notebook.
Generator to spatial work
Known cameras and synthetic truth made experiments repeatable, then led naturally toward shared coordinates and fused-view spatial reasoning.
Timeline
Research habits met working cameras.
The useful history is not a job-title list. It is the move from lab perception into hardware, real rooms, synthetic scenes, and eventually shared coordinates.
Joined Cortexica after the PhD period
The work moved from embodied AI research into computer vision under commercial constraints.
AI cameras under real constraints
Prototyping, edge inference, mobile ML, AR/VR experiments, and camera-as-computer objects.
Edge AI demos
Edge inference, public films, and practical objects that made vision legible in physical rooms.
Toward spatial intelligence
Synthetic-world tools and multi-camera thinking became a practical move toward spatial intelligence.
Working pattern
Physical constraints turned models into working vision.
The applied vision work kept the robotics habit of respecting hardware, lighting, latency, data gaps, and the need for working systems.
Move inference closer to the camera.
Make camera work legible in real settings.
Prototype quickly enough to discover what is real.
Use synthetic scenes when real data is missing or hard to label.
Representative scenes
Camera years under constraint.
Three scenes make the AI camera period easier to read as lived work.
CortexiCAM / Myriad-X prototype
The question moved into the camera body itself: limited compute, uneven light, latency, and a small edge device that still had to make perception visible without a lab machine beside it.
Watch CortexiCAMAxis Experience Centre demo
A physical room had to be readable to people, not just accurate in a notebook. The work was to turn camera output into visible states that made sense where the camera was being used.
Watch room-state filmGenerator and known camera truth
Some experiments needed a world where the camera positions and truth were known. Generator kept the work measurable when real data was scarce, expensive to label, or impossible to repeat exactly.
Watch GeneratorVideos
AI camera and edge vision.
The core objects show what this period contributed: hardware limits, edge inference, synthetic truth, and the bridge from cameras toward coordinates.
Overview
The Path of Innovation
AI camera demos from the Cortexica innovation period.
AI camera
CortexiCAM and Myriad-X edge AI
Myriad-X edge camera prototype in camera-as-computer form.
Room state
Axis Experience Centre demo
Entrance states made visible in a real room.
Article
UP Board edge-computing article
Guest article connecting GPU robotics, Movidius AI Core, UP Squared, and edge inference.
Generator
Synthetic worlds and known camera truth
Synthetic scenes with known cameras and controllable ground truth.
Spatial precursor
Six-view fused cloud precursor
Six calibrated angles forming one shared spatial object.
More camera-year filmsOpen two supporting room-state objects
Toward spatial intelligence
Known cameras make perception measurable.
The six-view fused-cloud precursor connects the camera work to later spatial calibration: multiple overlapping views treated as one physical scene.
Read spatial intelligence