Post-Storm Damage Assessment With Computer Vision

See how computer vision assists with storm damage assessments, from categorizing damage to creating maintenance tickets for repair crews. After a storm, crews need three answers before the first truck rolls: which poles are down, what broke on each one, and which crew goes first. Jennifer Kuchta and Alexei Alexandrovich, implementation engineers at Roboflow, show how one computer vision pipeline answers all three from drone footage, from labeling the training data to streaming live video from the drone. = Chapters = 0:00 Intro: Inspecting a Utility Pole by Drone 1:22 Demo: Find Downed Poles and Map Damage by GPS 3:15 Train for Any Class: Fallen Branches and Street Signs 5:29 How It Was Built: Labels From GPT-6 Astra and SAM 3 8:34 Why Train Your Own Model for Production 9:14 Train RF-DETR Segmentation in About 7 Minutes 10:17 Turn Detections Into Salesforce Cases With Workflows 11:29 Assign the Right Repair Crew to Each Finding 11:59 Stream a DJI Drone Over RTMP Into Roboflow 14:05 Bonus: Detect Leaning Poles From a Dash Cam = Resources = Auto Label with GPT-6 Astra: https://blog.roboflow.com/auto-label-with-gpt-6-astra/ RF-DETR in Roboflow: https://docs.roboflow.com/models/supported-models/rf-detr Join a future live session: https://roboflow.com/webinar