DIS icon indicating copy to clipboard operation
DIS copied to clipboard

Training with DUTS Datasets

Open skuley opened this issue 3 years ago • 1 comments

Hi, I've been training the model with DUTS datasets.

As you have mentioned in the paper, DUTS Images are relatively smaller sized than DIS5K Images.

I have customized layers as mentioned, but outputs are not satisfying.

My question is.. How can I improve output in such small sized datasets?

Also, if DIS is trained with 4K high resolution datasets, what happens when test with lower resolution than 4K?

Thanks.

skuley avatar May 23 '22 03:05 skuley

I feel that this model is specially created for objects that have a "latticed" structure and highly optimized for those. As you know each model is data dependent and it is evident in ablation studies that U2NET which was trained on DUTS, did not work better on this data.

Please let people know on this thread if you make breakthrough. I'm looking forward for some revelations.

deshwalmahesh avatar Jul 09 '22 14:07 deshwalmahesh