SushkoVadim
SushkoVadim
Hi, I am not sure there exists an official list of the train/val splits for the ADE20K-outdoor dataset. To construct this dataset, we selected images in the respective splits of...
Hi, As far as I could see in our evaluation scripts, we used the [official LPIPS repository](https://github.com/richzhang/PerceptualSimilarity) and used the VGG network: `lpips_net = PerceptualLoss(model='net-lin', net='vgg', use_gpu=True, gpu_ids=[0])` @edgarschnfld please...
Hi, Yes, this is possible, and it should be relatively little effort to implement a custom dataloader. Step 1: You should create a file `CustomDataset.py` in the `dataloaders` folder. Copy-paste...
Hi, Our project is mainly designed for semantic image synthesis, and we haven't tested it for general image-to-image translation. For example, the generator always expects to receive a semantic label...
Hi, yes opt.label_nc should be set to the number of semantic classes, and semantic_nc is set to `opt.label_nc+1` in case you have a "don't care", "unlabelled", or "other" label. So...
Hi, The input label map should be a one-hot encoding of shape [BxNxWxH], where B is the batch size, N is the number of classes in the dataset, WxH the...
Hi, First of all, a couple of questions: 1) Did you change the code after downloading the repo, so do you use exactly the version we released? 2) Do you...
Hi, Thanks for pointing out the inconsistency. That's a valid point. Indeed, without the *num_of_non_zeros_classes_in_batch* divider, the overall scale of the loss for the batch would be approximately proportional to...
Hi, We use the resolution equal to the size of generated images, which is 512x256 for cityscapes. The images should be taken from the `./results/oasis_cityscapes/best/image/,`, while the label maps should...
Hi, I think the easiest way to understand the folder structure and the label types is to have a look at one of the commonly used datasets. Taking Ade20k as...