Object detection product based on Nvidia Xavier NX , Yolo3 and Deepstream

After playing around with the Nvidia devices and software I though that its time to put it all together and attempt to make a saleable product.
I have made a Object detection unit based of the Xavier NX and Deepstream running yolo3. It has the capability to detect 79 different objects.
You can set count thresholds for objects and control relays with the results. Information gathered from the device can also be sent to a PLC for additional processing. It also has the capability to send e-mail alerts along with sending the info collected to a database to be analyzed.
Since the Xavier NX in the Object detecting unit is running as a server it can be accessed by any device on the same network. In addition the unit is streaming out a video signal that shows the state of the relays along with the objects detected in the video stream. This also can be accessed by any device on the same network.

3 Likes

Hi, this is very interesting!! What implementation of YoloV3 did you used?

1 Like

I am using the YOLO V3 model that is located at
“nvidia/deepstream/deepstream5.1/sources/objectDetector_yolo”

just a heads up.
There is a performance issue with this model in Deepstream 6.0.
Nvidia is working on it.

1 Like