AI-accelerated Defect Detection

CIGRE 2025 — IA for accelerated defect detection in UAV tower audits

32%

Time saved

42%

More defects

CIGRE 2025

Venue

PyTorchSAMMaxViTRT-DETRYOLO

First-author paper at CIGRE 2025 (Montreal). Demonstrates a 32% reduction in total inspection time and 42% more defects detected versus manual methods, using a hybrid SAM + MaxViT + RT-DETR pipeline trained on real UAV captures.

More Details

Extended write-up coming soon. Check back for a full breakdown of the pipeline architecture and experimental results.

Paper ↗