Hi,

I'm trying to generate images with the pretrained model and the provided preprocessed dataset, but I'm only getting random pixels. I wonder if I'm missing anything in my setup not mentioned in the README file. Really appreciate your help!

Sample output:

My test.sh: python test.py
--dataroot deepfashion
--dirSem deepfashion
--pairLst deepfashion/fashion-resize-pairs-test.csv
--checkpoints_dir ./checkpoints
--results_dir ./results
--name fashion_AdaGen_sty512_nres8_lre3_SS_fc_vgg_cxloss_ss_merge3
--model adgan
--phase test
--dataset_mode keypoint
--norm instance
--batchSize 1
--resize_or_crop no
--gpu_ids 0
--BP_input_nc 18
--no_flip
--which_model_netG ADGen
--which_epoch 800

My folder structure: ADGAN ├── checkpoints │   ├── fashion_AdaGen_sty512_nres8_lre3_SS_fc_vgg_cxloss_ss_merge3 │   │   ├── 1000_net_netG.pth │   │   ├── 800_net_netG.pth │   │   ├── loss_log.txt │   │   ├── opt.txt ├── cx ├── data ├── deepfashion │   ├── fashion-resize-annotation-test.csv │   ├── fashion-resize-annotation-train.csv │   ├── fashion-resize-pairs-test.csv │   ├── fashion-resize-pairs-train.csv │   ├── resized │   ├── semantic_merge2 │   ├── semantic_merge3 │   ├── test │   ├── testK │   ├── test.lst │   ├── train │   ├── trainK │   ├── train.lst │   ├── vgg19-dcbb9e9d.pth │   └── vgg_conv.pth ├── gif ├── losses ├── models ├── options ├── README.md ├── scripts ├── ssd_score ├── test.py ├── tool ├── train.py └── util

I also fixed a hardcoded path in model_adgen.py locally.

Hi @menyifang ,

How long did it roughly take to train the pretrained model with 2 V100 GPUs? A few days or weeks? (I read your paper but it doesn't seem to be mentioned there.)

I am trying to map the segmentation mask output with the merged (K=8) indexes. The current indexes I have are

np.array(('Background', # always index 0 'Hat', 'Hair', 'Glove', 'Sunglasses', 'UpperClothes', 'Dress', 'Coat', 'Socks', 'Pants', 'Jumpsuits', 'Scarf', 'Skirt', 'Face', 'Left-arm', 'Right-arm', 'Left-leg', 'Right-leg', 'Left-shoe', 'Right-shoe',)) is the input

and the merged index is : background, hair, face, upper clothes, pants, skirt, arm and leg

Is there a code you could share where this operation is performed? I am trying to reuse the pre-trained model

I am not able to reproduce the results using the pre-trained models.

The above is the output I am getting. Can you predict why I am getting this issue?

File "/home/user20202735/ADGAN/models/model_adgen.py", line 37, in forward style = self.enc_style(img_B, sem_B) File "/home/user20202735/.conda/envs/adgan/lib/python3.6/site-packages/torch/nn/modules/module.py", line 1051, in _call_impl return forward_call(*input, **kwargs) File "/home/user20202735/ADGAN/models/model_adgen.py", line 130, in forward xi = x.mul(semi)

Hi, thanks for your great work first!

I'm trying to reproduce your code, but I can not understand why these code using F.batch_norm in AdaptiveInstanceNorm2d, why not just F.instance_norm?

https://github.com/menyifang/ADGAN/blob/4dd70649ad136829b92dd6a1a823af7594a0220f/models/model_adgen.py#L355-L362

In your paper, the perceptual loss was:

# vggsubmod refers to certain layer of VGG.
Lper = L1(gram(vggsubmod(x)), gram(vggsubmod(y))  

However in your implementation: https://github.com/menyifang/ADGAN/blob/d948cb135801c83295e9427cab5d7d738436aa95/losses/L1_plus_perceptualLoss.py#L63-L72

Could you give some explanation on that ? thanks.

The path to /data/deepfashion/vgg19-dcbb9e9d.pth was hardcoded as your local machine path. So I made some changes to make it more dynamic as it was a pin in my ass when I tried running this model today. But honestly good work that you have done in this model. keep it up please <3.

After prepared the environment, then run the cmd "bash ./script/train.sh", I got the error like "RuntimeError: The size of tensor a (750) must match the size of tensor b (176) at non-singleton dimension 3", can you answer the question, Thank you very much!!!

I tested with bash python ./scripts/test.sh to test using pre-trained 800-netG model.

data is arranged as follows:

+—deepfashion
|   +—fashion_resize
|       +--train (files in 'train.lst')
|          +-- e.g. fashionMENDenimid0000008001_1front.jpg
|       +--test (files in 'test.lst')
|          +-- e.g. fashionMENDenimid0000056501_1front.jpg
|       +--trainK(keypoints of person images)
|          +-- e.g. fashionMENDenimid0000008001_1front.jpg.npy
|       +--testK
|          +-- e.g. fashionMENDenimid0000056501_1front.jpg.npy
|   +—semantic_merge
|   +—fashion-resize-pairs-train.csv
|   +—fashion-resize-pairs-test.csv
|   +—fashion-resize-annotation-pairs-train.csv
|   +—fashion-resize-annotation-pairs-test.csv
|   +—train.lst
|   +—test.lst
|   +—vgg19-dcbb9e9d.pth
|   +—vgg_conv.pth
... 

code reference

https://github.com/menyifang/ADGAN/blob/c76647172e923573b4012b6c17a1b3938155aedd/data/keypoint.py#L52:L88

I got following runtime error :

/ADGAN/data/keypoint.py", line 80, in __getitem__
 BP1 = BP1.transpose(2, 0) #c,w,h
 IndexError: Dimension out of range (expected to be in range of [-2, 1], but got 2)

debug output :

>>>BP1_img.shape
(256, 176)

Any suggestions how to solve this!

When I downloaded "In-shop Clothes Retrieval Benchmark", I got the following 9 error messages:

In-shop Clothes Retrieval Benchmark/README.txt

• You do not have permission to download this document. In-shop Clothes Retrieval Benchmark/Img/img.zip
• You do not have permission to download this document. In-shop Clothes Retrieval Benchmark/Anno/list_item_inshop.txt
• You do not have permission to download this document. In-shop Clothes Retrieval Benchmark/Anno/list_description_inshop.json
• You do not have permission to download this document. In-shop Clothes Retrieval Benchmark/Anno/list_landmarks_inshop.txt
• You do not have permission to download this document. In-shop Clothes Retrieval Benchmark/Eval/list_eval_partition.txt
• You do not have permission to download this document. In-shop Clothes Retrieval Benchmark/Anno/attributes/list_attr_cloth.txt
• You do not have permission to download this document. In-shop Clothes Retrieval Benchmark/Anno/list_bbox_inshop.txt
• You do not have permission to download this document. In-shop Clothes Retrieval Benchmark/Anno/attributes/list_attr_items.txt
• You do not have permission to download this document.

Is that OK to skip these files?

Hi, I am trying to download data (so many data...!!) anyway it said .ds_stre file is passwd protected and asked me passwd. Could you help me on this?

In fact, img_highres_seg-004 and img_highres-003 are passwd protected.

I'm running compute_coordinates.py in order to recalculate keypoints, but when I run it I have follow warning, (that I suppose mean some issue on size of images):

tensorflow:Model was constructed with shape Tensor("input_5:0", shape=(1, 368, 368, 3), dtype=float32) for input (1, 368, 368, 3), but it was re-called on a Tensor with incompatible shape (None, 184, 126, 3).

The script run, but it create a file with all -1 in keypoints.

What should be the issue?