Weird results about dis

[xuebinqin]

thanks for your interests guys, this version DIS dataset is trained with samples that human and animals are not included. We are preparing another model which will be released later to mitigate this issue. In addition, we will release our DIS v2.0, which will include highly accurately labeled humans, animals, etc. Please stay tuned. Thanks again for your interests.

…

On Tue, Jul 26, 2022 at 1:38 AM Swayam ***@***.***> wrote: Yeah it can be because of the huge diversity in dataset, that's why to excel in particular category I'm just re-training it again only on to that category. Also as mentioned in readme, pre-trained weights for commercial use are yet to be release, maybe they'll perform better. — Reply to this email directly, view it on GitHub <#19 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADSGORPLP6HEIYYRP6VMBQ3VV6PXBANCNFSM54UWSPQA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

-- Xuebin Qin PhD Department of Computing Science University of Alberta, Edmonton, AB, Canada Homepage: https://xuebinqin.github.io/

[SwayamInSync]

Yeah it can be because of the huge diversity in dataset, that's why to excel in particular category I'm just re-training it again only on to that category.

Also as mentioned in readme, pre-trained weights for commercial use are yet to be release, maybe they'll perform better.

[edgartaor]

Just for the record, this is the output with the DIS v2.0

[darwinOne]

@edgartaor which pretrained did you use to get this result, Sir?

[edgartaor]

DIS v2.0

[darwinOne]

@edgartaor Is it correct that this link is for the DIS v2 pretrained model?
https://drive.google.com/u/0/uc?id=1KyMpRjewZdyYfxHPYcd-ZbanIXtin0Sn&export=download

[VladMVLX]

DIS v2.0

Where did you get DIS v2.0 ? Can`t find any link

[Rijgersberg]

DIS v2.0

Where did you get DIS v2.0 ? Can`t find any link

Yes I was wondering that as well!

[edgartaor]

If I'm remember correctly, I downloaded from the link in the README.

** (2022-Aug.-17)** The optimized model for general use of our IS-Net is now released: isnet-general-use.pth (for general use) from (Google Drive) or (Baidu Pan 提取码:6jh2), please feel free to try it with the newly created simple inference.py code on your own datasets.

[VladMVLX]

I might be wrong, but pre trained model from README is DIS 1.0 general use