V2 ETA?

[haruntrkmn]

Great project. When will you approximately release the v2 dataset?

[RahulBhalley]

Hi @xuebinqin! The v2 demo results seems incredibly awesome!! Could you please tell us when will it be released?? Thanks.

[xuebinqin]

Thanks for your interest. Due to the heavy workload of the annotation, we can not give an exact ETA. It may take another few months.

On Mon, Oct 10, 2022 at 2:08 AM ਰਾਹੁਲ ਭੇਲੇ @.***> wrote:

Hi @xuebinqin https://github.com/xuebinqin! The v2 demo results seems incredibly awesome!! Could you please tell us when will it be released?? Thanks.

— Reply to this email directly, view it on GitHub https://github.com/xuebinqin/DIS/issues/44#issuecomment-1273009300, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADSGORI2IKWGIGEVQXKAM3LWCPMI3ANCNFSM6AAAAAAQMIOG5U . You are receiving this because you were mentioned.Message ID: @.***>

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

[RahulBhalley]

Thanks for a quick reply!

Maybe you can release the model trained on a subset of whole V2 dataset for now. I understand that's not a correct way of presenting research results but for production that result you demoed seems already a lot better than U2Net.

Again thanks for every other model you've released so far!

[SethRobinson]

I'm also looking forward to the updated dataset - and as a side note, I've implemented DIS into a game test (AI Paintball) that creates all the artwork "live" and it worked great. ( Video here )

[xuebinqin]

Thanks for your patience and interest. Our DIS V2.0 Dataset is almost there. But we need to finish the new paper for releasing that. It may take a while.

On Fri, Nov 11, 2022 at 1:10 PM Seth Robinson @.***> wrote:

I'm also looking forward to the updated dataset - and as a side note, I've implemented DIS into a game test (AI Paintball) that creates all the artwork "live" and it worked great. ( Video here https://www.youtube.com/watch?v=FoYY_90KlyE )

— Reply to this email directly, view it on GitHub https://github.com/xuebinqin/DIS/issues/44#issuecomment-1312194132, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADSGORKBLZENKOXVIFPTZHTWH2Y3ZANCNFSM6AAAAAAQMIOG5U . You are receiving this because you were mentioned.Message ID: @.***>

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

[RahulBhalley]

Interesting video @SethRobinson. So you basically segment an object and draw paint splash on it, right?

[SethRobinson]

Interesting video @SethRobinson. So you basically segment an object and draw paint splash on it, right?

Yes - I'd say the key thing is that I modified the text to image stable diffusion server it's using while the game is playing to have a "remove background" feature using DIS, so it returns the created image with an alpha mask cutout so the images are ready to be used as game sprites on the fly.

[RahulBhalley]

Wow! That's really an interesting use of Stable Diffusion but doesn't Stable Diffusion takes like a lot of seconds to generate just one paint splash image? I wonder how did you make it real time?

[SethRobinson]

Wow! That's really an interesting use of Stable Diffusion but doesn't Stable Diffusion takes like a lot of seconds to generate just one paint splash image? I wonder how did you make it real time?

It takes around 1.5 seconds to generate a single text to image on one nvidia 3090 graphics card - in my video, I'm using three Nvidia cards at once to run three servers (all on the same machine) so I can generate nearly two images a second.

Obviously this isn't something most people would have so this is all in the "just playing around with concepts" phase. :)

[dormal]

Thanks for your patience and interest. Our DIS V2.0 Dataset is almost there. But we need to finish the new paper for releasing that. It may take a while. … It's been six months, maybe it's time?

[polavishnu4444]

Thanks for your patience and interest. Our DIS V2.0 Dataset is almost there. But we need to finish the new paper for releasing that. It may take a while. … On Fri, Nov 11, 2022 at 1:10 PM Seth Robinson @.> wrote: I'm also looking forward to the updated dataset - and as a side note, I've implemented DIS into a game test (AI Paintball) that creates all the artwork "live" and it worked great. ( Video here https://www.youtube.com/watch?v=FoYY_90KlyE ) — Reply to this email directly, view it on GitHub <#44 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADSGORKBLZENKOXVIFPTZHTWH2Y3ZANCNFSM6AAAAAAQMIOG5U . You are receiving this because you were mentioned.Message ID: @.> -- Xuebin Qin PhD Department of Computing Science University of Alberta, Edmonton, AB, Canada Homepage: https://xuebinqin.github.io/

Any update?

[swframe]

I think you should create a patreon so we can support your efforts; I hope that might speed things up :)

[RahulBhalley]

Yeah. I hope you didn't abandon the project. @xuebinqin (When any project delays, TimberTron comes to my mind.)

[xuebinqin]

Thank you for your interest and support. We have had the v2 dataset ready and started the v2 model as well. The delay are mainly caused by the following three factors: (1) I changed my job and I can only work on this project in my spare time; (2) So that I need to re-configure all my hardwares and softwares from scratch; (3) More importantly, some of the V2 data samples are extremely complicated and these samples took a huge amount of time for labeling and polishing, hope you can understand the difficulties and challenges. Sorry for the delay, we are trying our best to release the new version as soon as possible.

On Mon, Mar 20, 2023 at 8:57 PM RahulBhalley @.***> wrote:

Yeah. I hope you didn't abandon the project. @xuebinqin https://github.com/xuebinqin (When any project delays, TimberTron comes to my mind.)

— Reply to this email directly, view it on GitHub https://github.com/xuebinqin/DIS/issues/44#issuecomment-1477246748, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADSGORLOCFKC3Z2452NNXF3W5ERLNANCNFSM6AAAAAAQMIOG5U . You are receiving this because you were mentioned.Message ID: @.***>

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

[VladMVLX]

any news?

[swframe]

@xuebinqin, Is there a tool you could recommend to help us segment our data? Thanks!

[RahulBhalley]

Hi @xuebinqin, could you please open-source your code and dataset (also the incomplete segmentations) so that we can help you do it??? We can also help train the model if you don't have time to re-configure your hardware. 🙏

[shkarupa-alex]

@xuebinqin will you release V2 dataset?