Skip to content
ZZNet in Detectron
Python CMake Cuda C++ MATLAB Dockerfile Makefile
Branch: master
Clone or download
Latest commit 8cea776 May 23, 2019
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
cmake Initial commit Mar 21, 2019
configs Initial commit Mar 21, 2019
demo Initial commit Mar 21, 2019
detectron fix mask May 14, 2019
docker
operator new op submit Mar 21, 2019
projects/GN Initial commit Mar 21, 2019
tools Initial commit Mar 21, 2019
.gitignore Initial commit Mar 21, 2019
CMakeLists.txt Initial commit Mar 21, 2019
CODE_OF_CONDUCT.md
CONTRIBUTING.md
FAQ.md Initial commit Mar 21, 2019
GETTING_STARTED.md Initial commit Mar 21, 2019
INSTALL.md
LICENSE Initial commit Mar 21, 2019
MODEL_ZOO.md Initial commit Mar 21, 2019
Makefile Initial commit Mar 21, 2019
NOTICE Initial commit Mar 21, 2019
README.md Update README.md May 23, 2019
requirements.txt Initial commit Mar 21, 2019
setup.py

README.md

ZigZagNet: Fusing Top-Down and Bottom-Up Context for Object Segmentation

by Di Lin, Dingguo Shen, Siting Shen,Yuanfeng Ji,Dani Lischinski,Daniel Cohen-Or,Hui Huang

Introduction

This repository re-implements ZigZagNet on the base of Detectron. Very consistent gains are available for all tested models, regardless of baseline strength.

Please follow Detectron on how to install and use ZigZagNet.

Citation

If you use our code/model/data, please cite our paper:

@inproceedings{cai18cascadercnn,
  author = {Di Lin, Dingguo Shen, Siting Shen,Yuanfeng Ji,Dani Lischinski,Daniel Cohen-Or,Hui Huang},
  Title = {ZigZagNet: Fusing Top-Down and Bottom-Up Context for Object Segmentation},
  booktitle = {CVPR},
  Year  = {2019}
}

and Detectron:

@misc{Detectron2018,
  author =       {Ross Girshick and Ilija Radosavovic and Georgia Gkioxari and
                  Piotr Doll\'{a}r and Kaiming He},
  title =        {Detectron},
  howpublished = {\url{https://github.com/facebookresearch/detectron}},
  year =         {2018}
}

Benchmarking

End-to-End Mask R-CNN Baselines

All models were tested on the coco_2017_val dataset

        backbone         type lr
schd
im/
gpu
box
AP
box
AP50
box
AP75
mask
AP
mask
AP50
mask
AP75
download links
X-101-64x4d-FPN-baseline Mask 1x 1 42.38 64.3 46.41 37.53 60.63 39.85 ——
X-101-64x4d-FPN-ZZNet Mask 1x 1 44.57 66.66 48.68 39.59 63.49 41.96 model
X-152-32x8d-FPN-baseline Mask 1x 1 45.30 66.85 49.85 39.75 63.54 42.25 ——
X-152-32x8d-FPN-ZZNet Mask 1x 1 46.44 67.69 51.15 40.96 64.88 43.66 model

Model

The model weights that can reproduce numbers in our paper will be coming soon.

You can’t perform that action at this time.