Computer vision · Academic and industry collaboration
Collaborative project with SEM Ltd.
Automated Transmission Tower Inspections
Drone imagery and deep learning for detecting transmission-tower components.
- Python
- PyTorch
- Computer vision
- Jetson Orin Nano
- DJI SDK
Overview and context
An AI-powered infrastructure inspection system built at Memorial University in collaboration with SEM Ltd. The system uses drone-based imaging and deep learning to detect transmission-tower components automatically. This was a team effort across academic and industry stakeholders — the work below is my own contribution, not sole ownership of the system.
Problem
Inspecting transmission towers manually is slow and difficult to scale; drone imagery is plentiful but needs automated interpretation.
My role and contribution
- Contributed to data preparation for drone-captured imagery.
- Worked on model training and evaluation.
- Contributed to edge deployment on NVIDIA Jetson Orin Nano hardware.
- Contributed to DJI drone integration and system-level validation.
- Collaborated with academic and industry stakeholders throughout.
Technical approach
- Drone-based imaging supplies the inspection dataset.
- Deep learning models are trained and evaluated for automated component detection.
- Inference runs on Jetson Orin Nano edge hardware rather than depending on an offboard server.
- System-level validation checks behaviour beyond individual model metrics.
Evidence and outcomes
- A working inspection pipeline from drone imagery through to detection on edge hardware.
- Validated integration between the DJI drone platform and the deployed models.