- 90% Theoretical Peak Performance
- 3x Increase in Per-Unit Computing Power
- 5ms Latency
Providing deep-learning-based object detection,tracking, recognition and attibute analysis intergrated back-end solution.
16-channel 1080p video streams
Full-scale real-time object detection and video
structuralization analysis with minimum object
size of 60x60 pixels
Providing deep learning-based object detection and tracking intergrated front-end solution.
1080p video stream
60x60 pixel specific target full detection
Advantages of "Nebula" and "Rainman"
16-channel real-time detection of single card
10 times energy efficiency ratio Increased