Need help with PANet?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

kaixin96
183 Stars 44 Forks 13 Commits 7 Opened issues

Description

Code for our ICCV 2019 paper PANet: Few-Shot Image Semantic Segmentation with Prototype Alignment

Services available

!
?

Need anything else?

Contributors list

No Data

PANet: Few-Shot Image Semantic Segmentation with Prototype Alignment

This repo contains code for our ICCV 2019 paper PANet: Few-Shot Image Semantic Segmentation with Prototype Alignment.

Dependencies

  • Python 3.6 +
  • PyTorch 1.0.1
  • torchvision 0.2.1
  • NumPy, SciPy, PIL
  • pycocotools
  • sacred 0.7.5
  • tqdm 4.32.2

Data Preparation for VOC Dataset

  1. Download

    SegmentationClassAug
    ,
    SegmentationObjectAug
    ,
    ScribbleAugAuto
    from here and put them under
    VOCdevkit/VOC2012
    .
  2. Download

    Segmentation
    from here and use it to replace
    VOCdevkit/VOC2012/ImageSets/Segmentation
    .

Usage

  1. Download the ImageNet-pretrained weights of VGG16 network from

    torchvision
    : https://download.pytorch.org/models/vgg16-397923af.pth and put it under
    PANet/pretrained_model
    folder.
  2. Change configuration via

    config.py
    , then train the model using
    python train.py
    or test the model using
    python test.py
    . You can use
    sacred
    features, e.g.
    python train.py with gpu_id=2
    .

Citation

Please consider citing our paper if the project helps your research. BibTeX reference is as follows.

@InProceedings{Wang_2019_ICCV,
author = {Wang, Kaixin and Liew, Jun Hao and Zou, Yingtian and Zhou, Daquan and Feng, Jiashi},
title = {PANet: Few-Shot Image Semantic Segmentation With Prototype Alignment},
booktitle = {The IEEE International Conference on Computer Vision (ICCV)},
month = {October},
year = {2019}
}

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.