Deep Learning-Based Perception Software

With compact, proprietary network available on 14 platforms, StradVision's SVNet drastically reduces network parameter size per frame computation, required memory, and power consumption. 

Also, SVNet algorithm has a proposal layer that makes it robust when perceiving objects in extreme weather conditions, small objects, or occluded objects, all with an extremely lean network.

Group 1
Group 2


graph_대지 1

Simultaneous ADAS & Driver Monitoring System

With one hardware and one network (SVNet)

Compact DNN-based Object Detection

Number of network parameters: < 2M

High Detection

Similar to faster R-CNN with VGG

Fully Optimized to the Target Platform

Real-time detection performance: 20-30 FPS


Your Hardware + Our Software = Fully Optimized ADAS Unit

SVNet can be fully optimized and deployed onto whatever hardware you wish to use for your ADAS and self-driving car needs in a very short timeframe.



StradVision's patented DNN, SVNet, is one of the few novel networks in the industry that meets the accuracy and computational requirements for commercial use on automotive embedded hardware. SVNet is applied into three sections dealing with external perception, internal perception, and tools to automate training and annotation.

SVNet External

Light, modular, and powerful perception algorithm already deployed in millions around the world from level 2 to 4 automation.​

SVNet Tools

By automating 97% of data annotation, our clients can aggressively scale up their data management capabilities while drastically improving the bottom line.


    How did you hear about us?

    StradVision's perception technology enables vehicles to recognize images in and out of the car, bringing about the advent of self-driving for the masses.