It's possible to use networks downloaded in "Hello AI world" in deepstream?


I was curious to know if the networks downloaded through the script provided for the “Hello ai world” can be executed in deepstream and how? each network has a .caffemodel, .prototxt and .json so i think that is possible but i dont know how to begin!


I think most of them are already included with the deepstream examples. You might have a look in


For models and sample videos and here …


… for various sample apps that use those samples

note: If you’re using deepstream-4.0 the path will be /opt/nvidia/deepstream/deepstream-4.0/...

If you want to use another model, it’s sometimes as simple as changing the filenames in the configuration .ini file. If you read the samples you willl understand. The nvinfer element (what does most of the work) reads a configuration file with parameters such as the model. Depending on what you want to do, you may not have to write any code at all. The reference deepstream_app is pretty flexible.

I tried the pednet linking the .engine .caffemodel and .prototxt of the network in the config file of sample 1 in deep stream. the example starts but with very low fps and also does not recognize anything