I’ve been enjoying the 2 days to a demo (taken about 5 days so far) and I want to say thanks.
It would be good if you could provide a more advanced pre-trained models for the different scenerios because it takes around 36 hours on my 1060 based GPU to train for dog detection.
I studied neural nets at uni over 10 years ago, it is amazing how far things have come. And to think back then people were sceptical about the benefits of neural nets being deeper than 2 layers deep.
Hi graham_chow, there are already pre-trained models provided for each scenario covered in the tutorial.
To use the pre-trained DetectNet models specifically, see these steps:
https://github.com/dusty-nv/jetson-inference#launching-with-a-pretrained-model
https://github.com/dusty-nv/jetson-inference#pretrained-detectnet-models-available
There are additional pre-trained models available which aren’t downloaded by default (to save disk space), you can uncomment them starting here.