Data checked 2026-08-13. Data may be stale - beyond the review cycle.
Review cycles are documented on the Methodology page. Methodology
microsoft/onnxruntime
Software Tier A edge_ai MIT
ONNX Runtime is a cross-platform inference engine that lets UAV developers deploy ONNX models across CPU, GPU and NPU backends with one format.
Engineering Snapshot
Use cases & tasks 7
- Best suited for
-
- Embedded AI engineers deploying models to flight controllers or companion computers
- Researchers exporting PyTorch/TensorFlow models to ONNX for cross-platform inference
- Teams needing hardware-accelerated inference on x86, ARM, NVIDIA Jetson, Qualcomm Snapdragon, or Intel OpenVINO targets
- Primary tasks
-
- Onboard object detection/tracking (e.g., YOLO, DETR exported to ONNX)
- Real-time segmentation for landing zone detection
- Obstacle avoidance inference pipelines
- Model serving in simulation-in-the-loop (AirSim, Gazebo) with hardware acceleration
Stack & ecosystem
- Resource type
- Software
- Ecosystem
- ai-framework · deep-learning · hardware-acceleration · machine-learning · neural-networks · onnx
License & compliance
- License
-
MIT(Inferred)
Lifecycle & freshness Show
- Maintenance
- Active
- Latest version
- Not recorded
- Last activity
- 2026-08-13
- Last checked
- 2026-08-13
- Verification
- Official confirmed
What It Solves
UAV developers need a single inference engine to deploy trained ONNX models across diverse onboard compute (CPU, GPU, NPU) without rewriting inference code for each backend.
Primary use cases
- Onboard object detection/tracking (e.g., YOLO, DETR exported to ONNX)
- Real-time segmentation for landing zone detection
- Obstacle avoidance inference pipelines
- Model serving in simulation-in-the-loop (AirSim, Gazebo) with hardware acceleration
Secondary use cases
- Benchmarking model latency across execution providers
- Converting scikit-learn pipelines to ONNX for edge deployment
- Federated learning model inference on swarm nodes
When to Use
Consider when
- Model is already in ONNX format or can be exported reliably
- Target hardware has a supported execution provider (CPU, CUDA, TensorRT, OpenVINO, QNN, CoreML, etc.)
- Deterministic latency and memory footprint are required
- License (MIT) permits commercial use without copyleft concerns
Verify before adopting
- Execution provider availability and maturity for specific SoC (e.g., QNN for Snapdragon, TensorRT for Jetson)
- Operator coverage for model opset version
- Dynamic vs static quantization support for target provider
- Threading/affinity configuration for real-time OS integration
- Binary size impact on constrained flash/storage
Start Here
Adoption Checklist
- Needs verification Execution provider availability and maturity for specific SoC (e.g., QNN for Snapdragon, TensorRT for Jetson)
- Needs verification Operator coverage for model opset version
- Needs verification Dynamic vs static quantization support for target provider
- Needs verification Threading/affinity configuration for real-time OS integration
- Needs verification Binary size impact on constrained flash/storage
Each check stays "needs verification" until an official source confirms it; unconfirmed items are never marked verified.
Known Limitations & Unknowns
Known limitations
- No built-in model training; inference-only
- Provider-specific operator support gaps (check compatibility per model)
- Large binary size when bundling multiple execution providers
- Real-time determinism depends on provider and OS scheduling
- NPU/accelerator support varies by vendor SDK version
Not publicly verified
- Exact latency/throughput on specific UAV companion computers (Jetson Orin, Snapdragon Flight, Raspberry Pi 5)
- Memory overhead of ORT C API vs Python bindings in long-running processes
- Thread-safety guarantees when called from multiple MAVLink/ROS2 callback threads
- Certification evidence (DO-178C, ISO 26262) for safety-critical deployments
Alternatives & Related Tools
Related tools
- tensorrt — Integrates with
How is it used?
Start from the recorded entry points below, then validate against the technical checklist.
- Official site https://onnxruntime.ai
- Repository https://github.com/microsoft/onnxruntime
- Documentation https://onnxruntime.ai
Technical checklist
- OK License identified Recorded: MIT
- OK Maintenance signal Active
- OK Verification status Official confirmed
- OK Source evidence attached 1 source record(s)
- NEEDS REVIEW Latest version recorded Not recorded
Official Links
Metadata & Governance
| License | MIT — Inferred |
|---|---|
| Commercial model | Unknown |
| Maintenance status | Active |
| Verification status | Official confirmed — Confirmed via the official repository API responses in SourceRefs below. |
| Latest version | Not recorded |
| Latest release | Not recorded |
| Last activity | 2026-08-13 |
| Last checked | 2026-08-13 |
| First seen | Not recorded |
Dataset facts
Facts above come from the official dataset card only; unconfirmed fields stay unknown.
Related Resources & Dependencies
- tensorrt — integrates with (confirmed)
Recent Activity
New resource: microsoft/onnxruntime
New repository resource added by the sprint promote pipeline.
Related Knowledge
Guides
- Deploying AI Inference on UAV Edge Compute — A repeatable workflow for choosing and validating an inference runtime for an onboard UAV computer.
Collections
- Edge AI Perception Starter — A starting stack for prototyping aerial perception models before edge deployment work.