https://github.com/nowsyn/SIM/blob/master/networks/model.py#L153

conv_out[-6]=conv_out[0]=https://github.com/nowsyn/SIM/blob/8a9f12ff92b2127c3988f799ba8878817e3f486c/networks/util.py#L176

so conv_out[-6][first 3 channels] is the input image. am i correct or am i missing something?

Hi,

Awesome work for introducing different forms of trimaps into training! After reading through the code implementation, there is one point I'm not sure about. In 3.1.2 'Simulating User Scribbles', it says 'To avoid the points being too close to each other, we set a threshold of 50 pixels between each two points'. After going through code in dataloader/Test_dataset/data_generator.py, class Genclickmap, seems there is no strict constraint applied on the distances between randomly selected fg points and bg points. I am not sure if I have missed some stuff, hope someone could help me. Thanks in advance.

I think I have correctly generated the test set with your code. but I got 29.58599 SAD with your weight from https://drive.google.com/file/d/1kmhQtO-6wXTxgHtQCLRj3xPOEtXyzTXC/view?usp=sharing and https://drive.google.com/file/d/12JCGqDylBXJpgDhj4hg_JZYdbHlX8TKe/view?usp=sharing. Fig. 8 row 1 in your paper.

Dear author ,

Thank you for your awesome work , but I am facing 1 issue whenever I try to download the classifier it gives me a error stating the file is corrupted , can u pls help me with this , Looking forward to hearing from you.

Best regards, Dane

1. I didn't find the training code, will it be released in the future?
2. I would like to ask whether this project is aimed at the matting of specific objects （Focus only on people, fire, trees... ）
3. For general object matting, is there any good implementation project, hope to recommend ，like this

hi, Here's my kaggle After I correct my sim.toml, I ran into the following error message: !python main.py -c config/sim.toml

2021-11-23 00:23:44,177-main.py:33-INFO-=> creating classifier 'resnet50' 2021-11-23 00:23:44,634-main.py:36-INFO-=> loading checkpoint 'checkpoints/classifier_resnet50_best.pth.tar' 2021-11-23 00:23:44,743-main.py:40-INFO-=> loaded checkpoint 'checkpoints/classifier_resnet50_best.pth.tar' modifying input layer to accept 11 channels 2021-11-23 00:23:45,403-main.py:63-INFO-Pretrain: no checkpoint found at 'None' 2021-11-23 00:23:45,404-main.py:66-INFO-Resume: loading 'checkpoints/SIM/ckpt_best.pth' Traceback (most recent call last): File "main.py", line 337, in main() File "main.py", line 327, in main model = build_sim_model(cfg.model, logger) File "main.py", line 67, in build_sim_model ckpt = torch.load(args.resume_checkpoint, map_location=torch.device('cpu')) File "/opt/conda/lib/python3.7/site-packages/torch/serialization.py", line 608, in load return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args) File "/opt/conda/lib/python3.7/site-packages/torch/serialization.py", line 777, in _legacy_load magic_number = pickle_module.load(f, **pickle_load_args) _pickle.UnpicklingError: invalid load key, '<'.

any help ?