I successfully reproduced your code on the atr dataset, but the effect did not reach the effect mentioned in your article, as shown in the figure below, epoch=50, batch=2, gpus=2, because batch=2 will report an error, I made a simple modification to the code, what else can I do to achieve the effect of your article. Also, which part of the code output should I look at, agm or npm, where mean miou is the human miou in the paper? Looking forward to your reply！

Hello Thanks for the awesome work

In the README, there are link to datasets that can be downloaded via google drive, can you specify what are the licenses for using these datasets

b.r

Hi, first of all, thanks a lot for sharing your work. Really impressive! I have 2 questions please:

1. I am not sure if I get the purpose/capabilities of this approach correctly but, is the model able to handle one single input image at inference time? e.g. during training it seems that both query and support images/masks are needed. For inference, what we want though, is to be able to submit a single (or batch) picture of a person and have the mask as output (image IN -> mask OUT). Is this supported?
2. If yes, do you have a code snippet providing preprocessing+inference+postprocessing?

Thanks and have a great day

Hello, I am running the project recently. I hope you can provide it. Please pass the application. My account number is [email protected] Thank you very much.

When I am running the project from the following command:

bash scripts/evaluate_atr_kwf1.sh

I am getting the following error:

Traceback (most recent call last):
  File "./exp/train.py", line 14, in <module>
    from torchvision import transforms
  File "/home/mayank/anaconda3/envs/iniy_oneshothumanparsing_041221/lib/python3.7/site-packages/torchvision/__init__.py", line 1, in <module>
    from torchvision import models
  File "/home/mayank/anaconda3/envs/iniy_oneshothumanparsing_041221/lib/python3.7/site-packages/torchvision/models/__init__.py", line 11, in <module>
    from . import detection
  File "/home/mayank/anaconda3/envs/iniy_oneshothumanparsing_041221/lib/python3.7/site-packages/torchvision/models/detection/__init__.py", line 1, in <module>
    from .faster_rcnn import *
  File "/home/mayank/anaconda3/envs/iniy_oneshothumanparsing_041221/lib/python3.7/site-packages/torchvision/models/detection/faster_rcnn.py", line 7, in <module>
    from torchvision.ops import misc as misc_nn_ops
  File "/home/mayank/anaconda3/envs/iniy_oneshothumanparsing_041221/lib/python3.7/site-packages/torchvision/ops/__init__.py", line 1, in <module>
    from .boxes import nms, box_iou
  File "/home/mayank/anaconda3/envs/iniy_oneshothumanparsing_041221/lib/python3.7/site-packages/torchvision/ops/boxes.py", line 2, in <module>
    from torchvision import _C
ImportError: libcudart.so.9.0: cannot open shared object file: No such file or directory

This is the structure of the project:

.
├── data
│   ├── datasets
│   │   ├── ATR_OS
│   │   │   ├── list
│   │   │   │   ├── meta_test_id.txt
│   │   │   │   └── meta_train_id.txt
│   │   │   └── support
│   │   │       ├── meta_test_atr_supports.pkl
│   │   │       └── meta_train_atr_supports.pkl
│   │   ├── CIHP_OS
│   │   │   └── list
│   │   │       ├── meta_test_id.txt
│   │   │       └── meta_train_id.txt
│   │   └── LIP_OS
│   │       └── list
│   │           ├── meta_test_id.txt
│   │           └── meta_train_id.txt
│   └── pretrained_model
│       └── deeplab_v3plus_v3.pth
├── dataloaders
│   ├── mypath_atr.py
│   ├── mypath_mhp.py
│   ├── mypath.py
│   ├── oshp_loader.py
│   └── transforms.py
├── exp
│   ├── __init__.py
│   └── train.py
├── how_to_install.txt
├── imgs
│   ├── Figure1.jpg
│   └── Figure1.pdf
├── LICENSE
├── networks
│   ├── deeplab_xception_synBN.py
│   ├── eopnet.py
│   ├── __init__.py
│   └── popnet.py
├── README.md
├── requirements.txt
├── scripts
│   ├── atr_eop_1wf1.sh
│   ├── atr_eop_1wf2.sh
│   ├── atr_eop_kwf1.sh
│   ├── atr_eop_kwf2.sh
│   ├── cihp_eop_kwf1.sh
│   ├── cihp_eop_kwf2.sh
│   ├── evaluate_atr_1wf1.sh
│   ├── evaluate_atr_kwf1.sh
│   ├── lip_etop_kw_f1.sh
│   └── lip_etop_kw_f2.sh
├── sync_batchnorm
│   ├── batchnorm.py
│   ├── comm.py
│   ├── __init__.py
│   ├── replicate.py
│   └── unittest.py
└── utils
    ├── __init__.py
    ├── metric.py
    ├── sampler.py
    ├── test_human.py
    └── util.py

Please help me.

bash scripts/evaluate_atr_kwf1.sh

the content of evaluate_atr_kwf1.sh： export PYTHONPATH="${PYTHONPATH}:$(pwd)" python ./exp/train.py
--batch 2
--gpus 1
--resume_model ./data/pretrained_model/atr_kwf1.pth
--resume_epoch 39
--lr 0.001
--numworker 6
--testInterval 10
--hidden_layers 256
--epochs 50
--dataset atr
--test_classes 18
--fold 1
--way 'kway'
--structure kway_oshp
--size 512
--prototype_warmup 25
--fg_weight 1.0
--contrast_weight 1.0
--testing_screening
--feature_lvl mid
--temperature 0.1
--validate

result: => Structure:kway_oshp, dataset:atr, way:kway, fold:1 => Saving log to: /home/fuzhi-dlj/workspace_big_bang/workspace/infant/human_parse/One-shot-Human-Parsing/exp/run_cihp/run_6 Constructing DeepLabv3+ model... Output stride: 16 Number of Input Channels: 3 => No model load! opts.resume_model ./data/pretrained_model/atr_kwf1.pth Traceback (most recent call last): File "./exp/train.py", line 560, in main(opts) File "./exp/train.py", line 381, in main net_.load_state_dict(x) File "/home/fuzhi-dlj/workspace/conda_list/pyt1736/lib/python3.6/site-packages/torch/nn/modules/module.py", line 1052, in load_state_dict self.class.name, "\n\t".join(error_msgs))) RuntimeError: Error(s) in loading state_dict for EOPNet_kway: size mismatch for prototype: copying a param with shape torch.Size([20, 256]) from checkpoint, the shape in current model is torch.Size([18, 256]).