SMIS - Semantically Multi-modal Image Synthesis(CVPR 2020)

Hello thanks for publishing this project. When I try to test: python3 test.py --name deepfashion_smis --dataset_mode deepfashion --dataroot /root/filters/SMIS/data/deepfashion --no_instance --gpu_ids 1 --ngf 160 --batchSize 4 --model smis --netG deepfashion

After placing the trained model in /checkpoints I get the following error:

Traceback (most recent call last): File "test.py", line 19, in dataloader = data.create_dataloader(opt) File "/root/filters/SMIS/data/init.py", line 44, in create_dataloader instance.initialize(opt) File "/root/filters/SMIS/data/pix2pix_dataset.py", line 22, in initialize label_paths, image_paths, instance_paths = self.get_paths(opt) File "/root/filters/SMIS/data/deepfashion_dataset.py", line 30, in get_paths label_paths_all = make_dataset(label_dir, recursive=False) File "/root/filters/SMIS/data/image_folder.py", line 49, in make_dataset assert os.path.isdir(dir) or os.path.islink(dir), '%s is not a valid directory' % dir AssertionError: /root/filters/SMIS/data/deepfashion/cihp_test_mask is not a valid directory

What am I doing wrong?

I use win10 pytorch 1.0.0, torchvision 0.2.2.post3, I try pytorch 1.4.0, torchvision 0.5.0 same error. I add DeepFashion Dataset and pretrained Run test: python test.py --name deepfashion_smis --dataset_mode deepfashion --dataroot D:/dataset/deepfashion --no_instance --gpu_ids 0 --ngf 160 --batchSize 4 --model smis --netE conv --netG deepfashion Error: Please support help me!

Hello, I downloaded the deepfashion dataset from your home page, but I didn't find the placement form of the dataset in your description document. It's somewhat difficult to understand and the following errors occurred，hope to get your help E:/codeproject/SMIS-master/train.py Traceback (most recent call last): File "E:/codeproject/SMIS-master/train.py", line 21, in dataloader = data.create_dataloader(opt) File "E:\codeproject\SMIS-master\data_init_.py", line 44, in create_dataloader instance.initialize(opt) File "E:\codeproject\SMIS-master\data\pix2pix_dataset.py", line 22, in initialize label_paths, image_paths, instance_paths = self.get_paths(opt) File "E:\codeproject\SMIS-master\data\coco_dataset.py", line 35, in get_paths label_paths = make_dataset(label_dir, recursive=False, read_cache=True) File "E:\codeproject\SMIS-master\data\image_folder.py", line 49, in make_dataset assert os.path.isdir(dir) or os.path.islink(dir), '%s is not a valid directory' % dir AssertionError: ./datasets/deepfashion/train_label is not a valid directory

can i make some new data for training ,beacause after reading the paper,i believe this network can fit into different application and i want have a try.

Hello！This is a good job.

I think that the new metrics (mCSD and mOCD) you designed is meaningful. But it seems that these new metrics are not included in the open source code. Are you willing to open the relevant code?

Looking forward to your reply！

Hi! I'm appreciate for the author‘s efforts. And i try the test, the cloth try on is nice,but the person's face is not good.

It seems like that the face restoration is not that good, is there any solution for this problem? As I think, use a mask that do not change the face? any suggestions? thx!!!

Hello！This is a good job.

I think that the new metrics (mCSD and mOCD) you designed is meaningful. But it seems that these new metrics are not included in the open source code. Are you willing to open the relevant code?

Looking forward to your reply！

Hello! Thank you for your wonderful project! When I trained ADE20k data set, some problems appeared. What is the value of-use _ vae10 in ade20k.sh? I set 10 to start training, but the speed is very slow. What equipment do you train on? About how long did you train? Thank you! Looking forward to your reply!

Hi @Seanseattle and the team. Congratulations and thank you for the wonderful work!

Using the file test.py, I have been able to generate multiple images in which the pose of the model comes from the input image, while the cloths are randomly generated. I would like to go a bit further by reproducing the test that you did in the paper where you take the cloth of one person an put it on another model. I couldn't find out how to do it with your source code. Could you please give me some instruction on this?

Thank you very much!

This paper is Awesome so I would request Plz include a Sample Notebook including custom data interpolation/Stylemix as documents are not clear. Greetings!