Classification of Astronomical Spectra Based on Multiscale Partial Convolution

Environment

• Ubuntu Server 22.04 LTS
• Python 3.10.8
• CUDA 11.7
• CUDNN 8.5

Create a new conda environment and install the required packages:

conda create -n mspc python=3.10
conda activate mspc
conda install pytorch torchvision torchaudio pytorch-cuda=11.7 -c pytorch -c nvidia
pip3 install opencv-python matplotlib scipy scikit-learn tqdm tensorboard tensorboardX torchinfo

Before training, check the cfg/cfg.py file to set your training configuration.

Data Download

All data can be downloaded from the SDSS and LAMOST official websites according to the fields in the data folder.

Dataset Directory Structure

Support K-fold cross-validation.

├── DATASET
│   ├── fold 1
│   │   ├── train
│   │   │   ├── xxx 1.csv
│   │   │   ├── xxx 2.csv
│   │   │   └── ...
│   │   ├── val
│   │   │   ├── yyy 1.csv
│   │   │   ├── yyy 2.csv
│   │   │   └── ...
│   │   ├── test
│   │   │   ├── zzz 1.csv
│   │   │   ├── zzz 2.csv
│   │   │   └── ...
│   ├── fold 2
│   │   ├── ...
│   ├── fold 3
│   │   ├── ...
└── ...

Training on other sky surveys

Please modify the read_other_fits function in data_preprocess/spec_preprecess.py to preprocess the data and build a dataset, and then modify cfg/cfg.py to fine-tune the model structure

Citation

@article{article,
author = {Wu, Jingjing and He, Yuchen and Wang, Wenyu and Qu, Meixia and Jiang, Bin and Zhang, Yanxia},
year = {2024},
month = {05},
pages = {260},
title = {Classification of Astronomical Spectra Based on Multiscale Partial Convolution},
volume = {167},
journal = {The Astronomical Journal},
doi = {10.3847/1538-3881/ad38ae}
}