Applied machine learning · 2025
Object Detection
An applied computer-vision project focused on the complete object-detection workflow rather than only model inference.
ROLE
Model training · Evaluation
STACK
PyTorch · Faster R-CNN · TorchMetrics · Python
OUTCOME
Covers the ML lifecycle from detection datasets and training to bounding-box inference and mAP evaluation.
Images + labelled bounding boxes
Faster R-CNN learns class + location
Confidence, IoU and mAP
PROBLEM
What needed solving
Object detection must identify both what an object is and where it appears in an image, which needs different targets and evaluation from simple classification.
APPROACH
How I approached it
I prepared detection data, trained Faster R-CNN, visualised predictions and evaluated model quality using detection metrics such as mean average precision.
ENGINEERING
Implementation highlights
ENGINEERING SUMMARY
What this project is built on.
TECHNOLOGY
PyTorch · Faster R-CNN · TorchMetrics · Python
ROLE
Model training · Evaluation
OUTCOME
Covers the ML lifecycle from detection datasets and training to bounding-box inference and mAP evaluation.
LINKS
NEXT PROJECT
Full-stack application
Compliance Tracker
A responsive compliance workspace for clients, tasks, deadlines, filtering and lightweight analytics.