Xuebin Qin

Sorry, what do you mean about too accurate? You can get your jpg image as model input and prepare your ground truth mask as binarized image with either pixel value...

please just follow the comments in the https://github.com/xuebinqin/DIS/blob/main/IS-Net/train_valid_inference_main.py in addition, you need to comment the 294 line out and uncomment the 295 line in " https://github.com/xuebinqin/DIS/blob/main/IS-Net/data_loader_cache.py". It is a temporary...

Thanks for your suggestion. We will think about it and probably provide a simple tutorial for the labeling skills, which are sometimes tricky but valuable to most of the users....

Thanks for your interest. DIS v2.0 is ongoing and haven't been released yet. On Tue, Aug 9, 2022 at 6:49 PM Sungyong Hong ***@***.***> wrote: > The link you tagged...

It is still ongoing. We need to submit the paper first and get accepted. It may take a few months. On Wed, Aug 10, 2022 at 6:18 PM Nguyễn Văn...

Hello, thanks for your interest. The model itself should be able to handle multiple object segmentation. But the output will still be binary and objects will not be separated. As...

Sorry for the confusion. Till now only academic version of DIS V1.0 has been released. We will later release another version of DIS V1.0 weights for general use. DIS V2.0...

please keep training after 2000 iterations, you can check the mae and Fmeasure. If they are in the reasonable range, e.g. F\belong[0.4,0.8], There should be an issue about the loss....

Thanks for your interests. Hi, Thanks for the interesting work. Just reviewing the paper here, I've got several questions regarding GT encoder. You described GT encoder to be self-supervised --...

Thank you for your interests guys. We will update our repo soon to include all the interesting applications. In addition, we are preparing the DIS V2.0 dataset. Please stay tuned....