6G-PATH OC2 [AI-DFS]: Patras Video-Inference Tests and Selected Local-Lab Measurements

This dataset contains 42 selected experimental cases from the AI-DFS (AI-DroneFireSense) use case within 6G-PATH: 14 clean-lifecycle Patras cases and 28 completed local-lab sampled observations. The experiments investigate recorded-video streaming and AI inference across cloud, edge and far-edge configurations, with the two evidence groups documented separately. The release includes per-case measurement records, receiver frame events and one-second frame-rate measurements where available, sampled worker telemetry, network-probe statistics, resource measurements, summary tables, methods, data dictionaries, selection records, SHA-256 checksums. This dataset-only release contains no Python scripts. It includes 161,552 normalized numeric points in the selected Influx exports. Original measurement timestamps and case boundaries are preserved. The local observations were collected on September 2026; the selected Patras cases span October 2026. Workloads include 720p, Full HD, 2.7K and UHD/4K profiles, with nominal 30/60 FPS variants where available. Original videos, model weights, packet captures and private deployment credentials are not included. Experiment completion is distinct from meeting performance targets: valid measured performance failures remain included. Local sampled worker rates are not frame-identity-verified receiver FPS, and local sample health does not establish continuous operation or clean shutdown. Reported application latency is measured between software endpoints, not physical camera-to-display latency. The dataset does not establish ten-stream capacity, long-term availability or detection accuracy. See the included methods and per-case records before comparing configurations.