Measured on SENSING Astra S36 Stereo Camera + Multiple Computing Platforms — Leading in Both Accuracy and Real-Time Performance
Ranging Accuracy (Comparison with Indoor Laser Rangefinder)
Test Environment: Jetson AGX Orin (32GB) + Astra S36 Camera, 24FPS, average of 10 measurements per distance, Actual Distance vs. Model Depth Value.
✎ Test Environment: Jetson AGX Orin (32GB) + Astra S36 Camera, 24FPS, average of 10 measurements per distance, Actual Distance vs. Model Depth Value.
Real-Time Inference Performance · Resource Usage
Environment: Indoor random scenes, continuous acquisition >3 minutes to calculate average values. S36 camera, resolution 640×480, different computing platforms and frame rate configurations.
✎Thor platform is limited by the camera's maximum frame rate of 30 FPS, and inference can reach 30 FPS; actual end-to-end latency is less than 5 ms. Under high load, GPU utilization on the Orin platform can reach 93%, while still maintaining stable 24 FPS output.
Differentiated Advantages · Built for Demanding Scenarios
For Algorithm Engineers and Decision-Makers — Delivering the Optimal Balance of Accuracy, Speed, and Reliability
Neural Fusion Architecture
Monocular Distillation + Transformer
Weak Texture / Reflective Surface
Robustness ↑41%
Native Camera Compatibility
Astra S56 · S36 Stereo
Factory-level Synchronized Calibration ·
Out-of-the-Box Ready
Super Real-Time Inference
Ultra-Low Latency <5ms
Structured Pruning Acceleration
Subpixel Accuracy
Relative Error ±0.9%
Occlusion / Texture-less Error ↓34%
Deep Empowerment · Full-Scenario Deployment
The sDepth + SENSING Stereo Camera solution has been widely applied in autonomous driving, robotics, and industrial inspection