Rail 04

Computer Vision & Perception

State-of-the-art algorithms for object detection, semantic segmentation, 3D reconstruction, and scene understanding — enabling machines to interpret the visual world.

Computer Vision & Perception

Computer Vision & Perception

Our computer vision research pushes the boundaries of visual intelligence — from real-time object detection in low-resource environments to full 3D scene reconstruction. We develop transformer-based architectures and lightweight inference pipelines suited to edge deployment in African contexts where compute is constrained.

YOLO Vision Transformers 3D Reconstruction Semantic Segmentation NeRF TensorRT

Key Research Topics

Object Detection

Real-time multi-class detection using YOLO-family and transformer-based detectors, optimised for low-power embedded systems.

Scene Segmentation

Pixel-level semantic and panoptic segmentation for autonomous driving, medical imaging, and satellite scene parsing.

3D Reconstruction

Neural Radiance Fields (NeRF) and photogrammetry pipelines for high-fidelity 3D scene and object reconstruction.

Edge Inference

Quantised, pruned models and TensorRT deployment strategies enabling full vision pipelines on edge and mobile hardware.

Current Projects

Vision · Active · 2024–2026

EdgeSight: Low-Power Detection Framework

Low-power object detection framework enabling real-time computer vision on edge devices with minimal energy consumption.

YOLO TensorRT ONNX C++ Python Raspberry Pi
Vision · Research · 2025

Scene3D: Neural Radiance Field Reconstruction

Neural radiance field reconstruction system for high-fidelity 3D scene modelling from 2D image collections.

NeRF 3D Gaussian Splatting PyTorch CUDA Open3D

Related Publications

Interested in Collaborating?

We welcome partnerships with institutions, NGOs, and researchers working at the frontier of AI and this domain.