Xnor’s DTx05 Person & Vehicle Detector for IP Cameras and NVRs

Learn More

Make any security system smart

Xnor’s DTx05 person and vehicle detector is a breakthrough in AI for commercial security. It’s so computationally efficient that you can run analytics on multiple video feeds using spare cycles on your existing NVRs. Through Xnor’s state-of-the-art deep learning models, DTx05 is able to run locally, on-device with no loss in accuracy or speed and a 10X reduction in power consumption and heat dissipation.

DTx05 gives you the metadata you need to create event notifications that are relevant and limit your reliance on professional monitoring. This reduces operating costs and allows you to quickly scale coverage. By running video analytics directly on IP cameras or local network appliances, privacy is protected and the need to push raw video to the cloud for analysis is eliminated.

DTx05’s efficient deep learning model enables customers to monitor up to 50 video feeds at 30 frames per second on one Intel Core i5 CPU with no GPU or hardware acceleration. This means 20% of that same CPU on an existing NVR can run analytics on 10 simultaneous feeds.


Fast person and vehicle detection with real-time streaming metadata delivers object class and bounding box coordinates right to your VMS or enterprise repositories.

Simple API for deploying Xnor object code on your cameras or NVR. Be up and running with just a few lines of code and easily update your models over the air.

Control memory and latency requirements for your system as well as object detection sensitivity via an intuitive thresholding algorithm.


Commercial Security Graphic

  • Eliminate the cloud and bandwidth costs of a continuous video stream by filtering what’s monitored and sending alerts for events involving people and vehicles.
  • Reduce hardware and server costs by deploying directly on the NVR or IP camera, eliminating the need for expensive cloud servers to run AI tasks.
  • Assure data privacy since the data never has to leave the premises to be analyzed. DTx05’s small memory and compute requirements allow inference to be done directly on-device.
  • Deploy fast and accurate state-of-the art deep learning models that detect a wide variety of people and vehicles.

Deployment Models and Example Benchmarks


On Device

Benchmark Tables Table 1


On a Network Appliance

On Network

Benchmark Tables Table 2


Contact Xnor to see how you can license the DTx05 for your security system


About Xnor

Xnor.ai brings ultra-efficient artificial intelligence capabilities to any platform, from resource-constrained edge devices like cameras, drones, handheld devices and IoT, to the most resource-rich large servers. Xnor’s proprietary technology enables even the most complex deep learning models to run on low-end, low-cost edge devices without any sacrifice in performance and efficiency. Advantages of Xnor technology include:

  • Data privacy - only metadata need leave the edge device
  • No high-bandwidth connectivity needed, as raw data can be processed on-device
  • 10X or more reduction in power consumption and heat dissipation
  • Vast improvements in latency

Xnor.ai is a spin-out of the Allen Institute for AI and the University of Washington. Its industry-leading machine learning technology is used by global enterprises in security, aerospace, industrial automation, automotive, retail, photography, consumer electronics and the public sector.