Jetson nano for object detection using mutiple camera inputs

Accelerated Computing Intelligent Video Analytics DeepStream SDK
,
1

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU): Jetson nano
• DeepStream Version: 5.0
• JetPack Version (valid for Jetson only):4.4
• TensorRT Version: 7.0

I am working on Jetson nano for an object detection and segmentation application. I’m trying Yolo and faster rcnn with resnet50 backend. I want to run yolo and faster rcnn on 4 camera inputs in real time with almost 25 classes. Is it possible to achieve that without much loss in FPS. Also, what will be the best pick as far as the algorithm(yolo and faster rcnn) and the backend network is concerned for jetson nano?

3

Hi,

You can find some benchmark result for Nano’s detector here:
https://developer.nvidia.com/embedded/jetson-nano-dl-inference-benchmarks

Due to the limited resource of Nano, we provide an object tracker inside our Deepstream SDK.
It can track the object in a short period so you don’t need to do the DNN inference every frame.

If this is acceptable, please check following topic for the detailed update to our YOLO sample in Deepstream.
It can reach 20fps with YOLOv3 on Nano, and it can be much faster with YOLOv3 Tiny.

Thanks.