How to use torch.utils.data.DataLoader for my custom dataset

vision
1

Sorry if this question is already asked.
I am new to the whole thing.
I want to know how to use DataLoader and split my data for training and testing sets.

So let’s say I have a folder called MyImgs, which includes 100 images, images are all 25x25 and they are RGB.
So my data is 100x3x25x25.
What should I do now, should I divide them into two separate folders (testing and training) and then do something or I don’t need to do that?

I am following the steps from the training your classifier tutorial, but I do not know how to change the lines that are related to loading the data.

Thank you very much for your help in advance and sorry for the naive question.

2

You can use TensorDataset to create dataset from a tensor. You can split the 100x30x25x25 tensor into two and create two datasets for train and val.

Creating your dataset is also fairly transparent. Check out this tutorial: http://pytorch.org/tutorials/beginner/data_loading_tutorial.html

1 Like