How to use a ready-trained neural network?

I found a neural network model that gets a depth map from an image in GitHub - OniroAI/MonoDepth-PyTorch: Unofficial implementation of Unsupervised Monocular Depth Estimation neural network MonoDepth in PyTorch. monodepth_resnet18_001.pth. How can I use this network? How can I just test it in my Pytorch?

You could read through the linked repository and see how the model is initialized and used. I would probably start by looking for a “main” or “train” function first. Once this is done, you could then try to rip out the model and use it in your own custom script.

BTW, DONOT forget torch.no_grad() and model.eval()

Thank you for your reply! I was busy for a while and could not continue my studies in this direction. Now I am here and would like to ask if there is already a more detailed discussion or article. Which says how can I use a ready-made neural network step by step?

Unfortunately, so far I don’t know what it says. :slight_smile:

Come on, there is even inference notebook attached :slight_smile:

I have an error with these lines, I don’t know if I need them. The error sounds like: unexpected token ‘%’

These are commands for IPython (if you’re running script using Jupyter)

thank you! I got it. :slight_smile: