Writing

Blog

Deep dives into AI, computer vision, and the engineering behind my research and projects.

Advanced June 15, 2025 18 min read

Instance Segmentation at Scale: YOLO, Mask R-CNN & SAM on HoloLens 2 Data

A deep comparative study of three state-of-the-art segmentation architectures applied to a real mixed-reality dataset — with pipeline design, training tricks, and honest benchmark results.

YOLOMask R-CNNSAMInstance SegmentationHoloLens 2

Abdoul-Wahabou H. Tiambou

MSc Computer Science · AI/ML & Computer Vision

From 2D Masks to 3D Meshes: Object Reconstruction with OccupancyNet on HoloLens 2

Featured

Advanced22 min read

From 2D Masks to 3D Meshes: Object Reconstruction with OccupancyNet on HoloLens 2

OccupancyNet learns a continuous occupancy function over 3D space instead of predicting discrete voxels. I explore how we used it to reconstruct 200+ indoor objects from partial HoloLens 2 observations — and what the limits are.

July 10, 2025Read

Vision Transformers Explained: From ViT to DINOv2 — A Practitioner's Guide

Featured

Intermediate25 min read

Vision Transformers Explained: From ViT to DINOv2 — A Practitioner's Guide

Transformers have conquered computer vision. In this guide, I break down the fundamental mechanics of ViT, trace its evolution through to Meta's DINOv2, and share practical code for leveraging frozen DINOv2 features for downstream tasks.

August 5, 2025Read