Weakly Supervised Semantic Segmentation for Join Key Local Structure Localization and Classification of Aurora Image
By Chuang Niu, Jun Zhang, Qian Wang, and Jimin Liang
A weakly supervised semantic segmentation method for joint KLS localization and classification of aurora image is implemented in this project, and the related paper is submitted to TGRS. More details will be descibed.
This project is based on:
It is noted that the original codes of selective search and fast-rcnn are not directly uesd by this project, but you must make sure that they can run normally before implementation of this project.
- Get the code. We will call the directory that you cloned Aurora-ASI-KLS into
$KLS_ROOT
git clone https://github.com/niuchuangnn/Aurora-ASI-KLS
cd $KLS_ROOT/selective_search_py
wget http:https://cs.brown.edu/~pff/segment/segment.zip; unzip segment.zip; rm segment.zip
cmake .
make
cd &KLS_ROOT/fast-rcnn/caffe-fast-rcnn
make all
make pycaffe
cd &KLS_ROOT/fast-rcnn/lib
make
- Download the region detection model: vcc_cnn_m_fast_rcnn_b500_iter_10000.caffemodel
cd $KLS_ROOT/Data
mkdir -p region_classification/output
Put the downloaded model into this folder.
- Run demo.
cd $KLS_ROOT/src/demo
python demo.py
You will see:
classification time: 1.2338631897
segmentation time: 1.77530801296
