Kenosis-quantized INT8 models verified against COCO ground truth — 93.6% of FP32 mAP retained at a fraction of the compute. Measured detection accuracy, not proxy metrics.
SPSC ring buffers paired with zero-copy frame transport. Strict per-camera thread isolation with zero dropped frames at real-time tiers — counter-verified by per-ring telemetry, not assumed.
Built on a memory-safe Native Rust Pipeline. We orchestrate industry-standard execution engines without the Python overhead, container bloat, or unpredictable runtime pauses.
Intelligent execution provider routing across CPU, CUDA, or DirectML layers. Maximizing throughput regardless of the underlying host environment.
Prometheus and REST monitoring with per-stage p50/p95/p99 latency histograms, per-ring-buffer fill and drop accounting, and overload alerting. Every performance claim on this page is reproducible from the daemon's own counters.
Measured — dual-camera headless daemon, re-validated 2026-07-13
2× 640×480@30 webcams · PP-YOLOE+ S INT8 @320 · Intel i5-13420H · ONNX Runtime CPU · ≥115 s steady-state windows, zero frame drops