Using image-classification pre-trained models only for inference


I want to do the below steps for image classification

1] To download the pre-trained model like resent, vgg etc…
2] Use it only prediction on my own dataset.
3] Also, use it for feature extraction. Is it possible to use TLT pre-trained models to extract features.

Also, when I download and unzip any image classification pretrained models, only “cspdarknet_19.hdf5” is present all the time. Does it mean like all the models will be present in this file itself?

You can use the TLT pretrained models(hdf5 file) to train your own dataset.
But cannot use it to run inference directly. You can train a tlt model. Then the tlt model can be used to run inference.