Due to active galaxy nuclei (AGN) mechanism at the center of the galaxies, electronics of high energy are ejected to blow and push the gas around. Bubbles or cavities are then generated, which can be detected at multiple frequency bands, especially for the X-ray band.
Since AGN reveals quite lots of attracting physical phenomena, detecting of them is significant. However, there exist many difficulties in our works. For instance, the background and system noise in the X-ray images, which lead to low Signal-to-Noise Ratio (SNR), should be eliminated. In addition, the high brightness (temperature) of the gas in the central region usually leads to low contrast of the cavity regions.
In this repository, we provide a toolbox namely cav_gcnn to detect cavities in the X-ray astronomical images observed by the Chandra X-ray Observatory (CXO). Our method is designed based on the state-of-the-art Convolutional Neural Network (CNN), as well as a strategy to handle the imbalanced dataset namely granularization.
To utilize our toolbox on cavity detections, a Granular CNN (GCNN) model should be trained in advance and saved. Then, cavities in the new observations can be detected and marked with elliptical markers after preprocessing on the raw image data. In this work, 40 observations of 40 different objects were applied to train our GCNN classifiers, and a snap of them is illustrated.
If you want to see the details of preprocessing and usage of our script, please refer to our paper and the python codes. And the installation of the toolbox is as follows,
$ cd cav_gcnn
$ <sudo> pip install <--user> . To run our scripts, some python packages are required, which are listed as follows.
The requirements file is provided in this repository, by which the required packages can be installed easily. We advice the users to configure these packages in a virtual environment.
- initialize env
$ <sudo> pip install virtualenv
$ cd cav_gcnn
$ virtualenv ./env- install required packages
$ cd cav_gcnn
$ env/bin/pip install -r ./requirements.txt- install the latest Theano, and Lasagne
$ git clone https://github.com/Lasagne/Lasagne.git
$ cd Lasagne
$ <sudo> pip install <--user> -e .$ git clone git://github.com/Theano/Theano.git
$ cd Theano
$ <sudo> pip install <--user> <--no-deps> -e .In addition, the computation can be accelerated by parallelly processing with GPUs. In this work, our scripts are written under the guide of Nvidia CUDA, thus the Nvidia GPU hardware is also required.
In addition to the GCNN package, we also provide two executable script for you to train the network, and detect cavities on new observations. The usages are as follows,
- Training
$ gcnn_train <inpath> <outpath> <numepoch>- Detection
$ gcnn_detect <obspath> <netpath> <numgra> <matname>It should be noticed that the matname for the detection should be in such structure like,
sample_<boxsize>_<overlaps>.mat
More details of the usages can be referred by keyword --help after the commands.
- Zhixian MA <
zxma_sjtu(at)qq.com>
Unless otherwise declared:
- Codes developed are distributed under the MIT license;
- Documentations and products generated are distributed under the Creative Commons Attribution 3.0 license;
- Third-party codes and products used are distributed under their own licenses.
This work has been accepted and published, which can be cited under the bibtex style as follows,
@article{Ma2017cavity,
title={An approach to detect cavities in X-ray astronomical images using granular convolutional neural networks},
author={Ma, Z., Zhu, J., Li, W., and Xu, H.},
journal={IEICE TRANSACTIONS on Information and Systems},
volume={100},
number={10},
pages={2578--2586},
year={2017},
doi={10.1587/transinf.2017EDP7079},
publisher={The Institute of Electronics, Information and Communication Engineers}
}