C3W2B_Assignment Pix2Pix Error In Given Code?

Course Q&A Generative Adversarial Networks (GANS) Apply Generative Adversarial Networks
,
1

image

transform = transforms.Compose([
    transforms.ToTensor(),
])

import torchvision
dataset = torchvision.datasets.ImageFolder("maps", transform=transform)

def train(save_model=False):
    dataloader = DataLoader(dataset, batch_size=batch_size, shuffle=True)

From this code snippet, it seems that the code is loading images from a folder named “maps” and this folder contains subfolders for both training and validation. Isn’t this wrong, shouldn’t it just be from the train folder?

2

Training and validation happen both during training phase, so thats fine!

3

The training dataset is used to train the model’s parameters, the validation dataset is used to fine-tune hyperparameters and evaluate model performance during training, and the test dataset is used for evaluating the model’s performance on unseen data to gauge its generalization ability.

4

@JOSE_Paul, you’re right that in this assignment, we’re just looking at both the “train” and the “val” images together. The dataloader considers the “train” and “val” sub-folders to be classes, but as you probably noticed, in the code, we are just ignoring the second return result (the class) when we call the dataloader.

        for image, _ in tqdm(dataloader):

This is OK for this assignment, since the point of the assignment is to just focus on implementing the loss function, and we just need a set of images to work on. We don’t go through a separate validation step in this assignment, but just visually look at the images to get an idea of how our fake images compare with their real counterparts.

1 Like