Speaking

Taxonomy over volume

How you define the classes a vision system looks for matters more than model size or data volume. Built from three years of segmentation work, and the subject of a paper in preparation.

From floorplan to 3D model

What it takes to turn a drawing into a structure a machine can use, and where the hard cases are. Based on the Akiya3D pipeline and its 13-class segmentation taxonomy.

Computer vision on a production line

Real-time inspection with no neural network at all, on a Raspberry Pi talking to a PLC. When classical methods are the right answer, and how to prove they work.

Making a photographed space queryable

Photographs to a 3D scene you can ask questions about: reconstruction, fusing 2D detections into 3D positions, and indexing what is where.

Tackling Japanese Traditional Floorplans: A Lightweight Segmentation Pipeline

Accepted

Enhancing Agricultural Supply Chain Traceability with Blockchain, Smart Contracts, and E-Labelling

Presented online

Leveraging Machine Learning Techniques in Converting 2D Floorplans Images to 3D Models

Presented online. Best Paper Award

AIOBI, Ouagadougou, Burkina Faso

Designed and delivered a computer vision training programme, beginner to intermediate. Built the full teaching materials and adapted the projects to local context.

Burkina Institute of Technology

Taught programming and introductory computer science to 120 Bachelor-level students over two academic years, alongside running student affairs.

AI2M4RI-2026 (Elsevier)

Reviewer

JoVE Journal

Reviewer, meta-ensemble deep learning for floorplan analysis