Question about Mask Images' Shape and Cropping for PIPE Dataset
#4
by
QinmingZhou
- opened
Hello, your dataset has been of great help to me. However, I would like to ask why the mask images you uploaded come in different shapes, and how you cropped them to perfectly match the dataset in PIPE. I would greatly appreciate it if you could share your insights. Thank you.
Hi!
Thanks so much for your interest — we’re happy to hear the dataset has been helpful to you!
All the images in the main dataset were resized using torchvision.transforms.Resize((512, 512)).
If you apply the same resize operation to the mask images, their shapes should match perfectly with the corresponding images in the PIPE dataset.
Hope this helps!