AI Powered Xiaomi Fang (IP cam)

6 0
  • 0 Collaborators

In this project we will connect an Intel Movidius Neural Compute Stick to a very popular and inexpensive IPcam (Xiaomi Fang). This will enable us to process the images directly on the camera to offer object or person recognition. ...learn more

Project status: Under Development

Artificial Intelligence

Groups
Student Developers for AI

Intel Technologies
Movidius NCS

Links [1]

Overview / Usage

In this project we will connect an Intel Movidius Neural Compute Stick to a very popular and inexpensive IPcam (Xiaomi Fang).
This will enable us to process the images directly on the camera to offer object or person recognition.
With this feature we will not only able to stream the video of the camera but also information about what it is seeing. It could be used to fire some triggers or to notify the user when some specific events happen.

While similar feature are already exposed by the camera, the opportunity to use the Intel Movidius Neural Compute Stick dramatically improves the accuracy and the capabilities of the device.

Methodology / Approach

In order to proceed with the project we must perform two steps:

The first thing to do is to connect the data lines of the usb port on the camera. The exposed usb port, in facts, only provides power to create an array of cameras.

To do this we have to remove the wifi card that is soldered on the main board and expose the usb data connection. Then we can solder a simple usb hub, reconnect the wifi card (to the hub this time) and use one of the hub ports to provide a data line to the usb socket.

Then we have to crosscompile the libraries and the application that will run on the device. This part is still work in progress and will be updated soon

Comments (0)