Xuebin Qin

Thanks for your interest. It depends on the dataset resolution. Larger input size will help with retraining details of those images with larger sizes. In our next paper, we will...

Sorry, what do you mean about the big-O? The running time of the ablation is provided in Table 2: https://xuebinqin.github.io/U2Net_PR_2020.pdf On Thu, Jul 21, 2022 at 12:52 PM Benny ***@***.***>...

you mean portrait generation or human portrait segmentation. For portrait generation we trained with 512x512 input size. Other settings are kept the same with U2Net for SOD. The only issues...

You can either flip the white and black in training stage or do that in inference stage. On Tue, Jul 26, 2022 at 8:55 PM zozoteacher ***@***.***> wrote: > Thank...

Theoretically the ground truth mask should only contain 0 and 1 or (0 and 255). You can threshold them with 0.5 and 128 or just keep them. It won't make...

Hi, Xiong, I think you can try bce first. According to my experience, bce usually works well because it often generates relatively smooth boundaries. Of course, it depends on your...

Yes, there is no change. On Sat, Jan 8, 2022 at 10:06 AM Testhjf ***@***.***> wrote: > @xuebinqin hi,I have a quertion ? the > trained u2net_human_seg.pth is based on...

you need to double check your code for making sure that your data is correctly loaded. The simplest way is to overfitting a one or two sample images. On Fri,...

hi, I think you can try u2net (human segmentation model, for human overall shape segmentation) + pymatting (https://github.com/pymatting/pymatting, for getting hair matting) to achieve an relatively high quality alpha channel...

or you can use rmbg directly: https://github.com/danielgatis/rembg, to see if that satisfies your requirements. On Tue, Apr 12, 2022 at 6:39 PM Xuebin Qin ***@***.***> wrote: > hi, > >...