Next-gen Dynamic UAV using Power9 Systems
summary
A drone system classifies fire vs non‑fire images using a convolutional neural network trained on a large mixed dataset from forestry departments and public sources, then deploys the model on a multirotor with thermal and RGB cameras (including 30x optical zoom), PDB, ESCs and a flight controller for stabilized flight and real‑time alerts. Training and optimization run on Power9 AC922 servers (NVLINK/PCIe Gen4/OpenCAPI) claiming up to 5.6× bandwidth; field detections are checked against that year’s fires by domain experts.
AI-written summary · about
chapters (7)
- 0:00Introduction to UAV project
- 7:12UAV hardware components overview
- 13:01Flight control systems explained
- 21:34Power management in drones
- 29:15AI in UAV applications
- 37:15Business model for UAV tech
- 44:12Summary and Q&A
related talks
30mUAV-based Person Detection and Geolocation for Search and Rescue
53mEfficient Visual Avoidance and Control for UAVs with Embedded GPUs
45m(Deep) Learning to Fly
48m(Deep) Learning to Fly
30mAI On-board of Satellites, Towards Autonomous Scientific Instruments
16mlifesaving drones that mine XMR
28mGame of Drones
21mVisualizing robot minds, and other fun stuff