About data split (train/validation)

Question

About data split (train/validation)

wooseok-shin opened this issue 3 years ago · 4 comments

Hi,
I have a question about the experimental setup.

In PraNet paper,
"the images from Kvasir, and CVC-ClinicDB are randomly split into 80% for training, 10% for validation, and 10% for testing."

Contrary to what is mentioned in PraNet paper, there seems to be no validation in the implemented code.
Did you actually train without validation? Or am I missing something?

Thank you.

DengPingFan commented 3 years ago

Yes.

Answer 1 · 2022-02-04T12:22:18.000Z

Yes, we train this model without validation. And we use the early-stop strategy.

Best,

Answer 2 · 2022-02-04T16:05:51.000Z

Thanks for your reply.

I have one more question,

Checking the number of TrainDataset.zip files you provide, it seems that the Kvasir and CVC-ClinicDB datasets are used for training together, is that correct?

Answer 3 · 2022-02-04T17:03:43.000Z

Thanks for your reply.

In PraNet paper, "the images from Kvasir, and CVC-ClinicDB are randomly split into 80% for training, 10% for validation, and 10% for testing."

In PraNet paper,
"the images from Kvasir, and CVC-ClinicDB are randomly split into 80% for training, 10% for validation, and 10% for testing."