We use Jetson AGX SOMs and Deepstream for video analysis at the edge. Some of our cameras, although not attached to a car, are somewhat low to the ground and can be blind from time to time.
I am interested to see if there is a way to use something like ClearsightNet from Drive in Deepstream. Can anyone point me in the right direction?
If you have got the ClearSightNet model. It can be integrated to DeepStream 6.0 GA.
If the preprocessing and inferencing are the same to resnet10, you can integrate with gst-nvinfer plugin Gst-nvinfer — DeepStream 6.1.1 Release documentation and you can implement your own post-processing with enabling gst-nvinfer “tensor-meta-output”.