Columns:

The train.csv file must have a column named "label" representing the true digit label for each image.
The test.csv file must have an additional column named "id" to uniquely identify each test image.
Pixel Values:

Pixel values for each image are provided in separate columns following the label column in train.csv and without label column in test.csv.
Pixel values should be used for model training and prediction. The image is assumed to be square, with equal height and width.