Oil & Gas Leak Computer Vision
A computer-vision pipeline for oil and gas leak detection built during a four-year AI research fellowship.
Industrial leak detection required an applied imaging pipeline capable of extracting a reliable signal from a large visual dataset.
Designed and evaluated a computer-vision pipeline across more than 10,000 images.
Lead research fellow, model development, experimentation, evaluation, and publication.
Research scope
The project explored computer imaging for oil and gas leak detection. It combined data preparation, model experimentation, evaluation, and research communication in a multi-year fellowship environment.
Evidence
- More than 10,000 images processed by the pipeline.
- 94% reported precision.
- Peer-reviewed AI and computer-vision research presented at conferences.
- Work completed as part of a fellowship awarded to a top-5% Informatics student.
Foundation for production work
The research built the evaluation discipline that later informed production ML work: define the metric, understand the failure mode, and treat model output as one part of a larger system.