How to pass a RGB dataframe to CNN?

I have a dataframe of (224 * 224 * 3) columns representing the pixel values of image and last column is the binary classification label. How do i proceed with this data to pass it to a CNN

To do so, better way would be to first convert the dataframe to numpy 3d array, and then use torch.from_numpy().
An even better approach would be to directly read csv file as numpy array instead of dataframe, and then convert it to torch.

1 Like

hi @Ishvinder , after following your instructions, now i have a tensor of images and a list of label values. How i can use them with dataloader for passing it to cnn or is there any other method without it

Hi @Rishabh,
Once you have all your data into a tensor, and you want to convert it into a DataLoader if I’m not wrong…

To do so, let’s say you have two tensors, one containing the labels, and other having the image data.

from torch.utils.data import TensorDataset, DataLoader
dataset = TensorDataset​(​tensor_x​,​tensor_y​)
dataloader = DataLoader(my_dataset) 
1 Like