Is it possible to save the preprocessed images from deepstream. Second, how can I disable the preprocessing in deepstream so I can feed the preprocessed images saved from tao deploy to the trt engine?
What kind of preprocessing are you mentioning?
You cannot do that. If you just want to use the TensorRT without deepstream, you need to implement that yourself.
In the config file it is necessary to add this:
net-scale-factor=1.0
offsets=103.53;116.280;123.675
model-color-format=1
I want to feed to the deepstream pipeline with already preprocessed images, so is there a way to do that?
Yes. You can use nvdspreprocess to prepare your own tensor data. You can learn how to use that through the guide and our demos like sources\apps\sample_apps\deepstream-preprocess-test.