AI Inventory

AI Inventory

One Deep Runner Z4I automatically counts the number of designated objects from each of a large number of cameras. By applying a high-recognition rate deep learning algorithm to high-resolution images, it automatically recognizes the presence and number of designated objects as well as their location on a regular basis and reports them to your IT system. For example, you can automatically manage how many products are on a warehouse shelf, which products are sold out and empty on a shelf, and how many containers are in a container yard.


  • Automatic inventory management
  • Automatic detection of empty shelves in the store
  • Check container shipment status
  • Identify the number of cars in the parking lot
< Automatic inventory management >
< Automatic inventory management >
< Container shipment status >
< Container shipment status >
< Store merchandise management >
< Store merchandise management >


  • It has a built-in IP camera interworking interface to connect low-cost, high-quality CCTV cameras.

  • It can also be connected to an existing CCTV system.

  • Up to 32 cameras can be recognized by one Deep Runner device.

  • By supporting high-resolution object recognition, even small objects on the screen can be accurately counted. To do this, we use our proprietary technology to iteratively apply a low-resolution SSD object recognition algorithm to different locations and sizes of an image and incorporate the results.

  • We provide a deep learning training software environment for registering new products. Using the images of the new objects, you can simply train it to recognize. In addition, we provide technical support upon request.

  • Because it embeds a web protocol, it can easily link with a remote server to report recognition results and manage settings remotely.

  • It recognizes the registered cameras sequentially and it takes about 3 seconds to recognize one camera.

  • You can build the system at a low cost.

<AI Inventory Connection Diagram>

< Examples of warehouse management >
< Examples of warehouse management >

Based chipset

  • Xilinx Zynq UltraScale+ MPSoC


  • Remote control functions

            New product registration for each camera (deep learning model registration)

            RTSP address mapping per camera

            Registration of exclusion zone for each camera

            Camera video remote monitoring

  • Real-time output

            Count value by product type by camera

            Location coordinates by object