This repository contains scripts to be run at the Mauá server for the production phase of the star/quasar/galaxy classification for S-PLUS DR5. The training process and performance analyses are in another repository to be released within further notice.
- Environment:
- Python 3.10.12
- All packages were updated to the latest version for Python 3.10.12 (2021/10/21). Check file environment.yml for details.
- A new script named sqgtool.py was added. This script is the main script to be run on the Mauá server. It parses command-line arguments and calls the other scripts.
- A better logging process was added. Logs are stored in /logs/.
- Multithreading can be used for both crossmatch and classification processes. The number of threads can be set by the user.
The classification for S-PLUS DR2 and DR3 follows Nakazono et al. 2021. The classification for S-PLUS DR4 follows the same procedure as DR2/DR3 but with updated performance (as seen in the documentation) due to changes in reduction and calibration processes.
The S-PLUS DR5 has further improvements in reduction and calibration processes. We made a few improvements in the classification process, briefly described below:
- [Base model] Re-trained the classification model that uses S-PLUS magnitudes + WISE magnitudes + morphological parameters using the new data
- The model mentioned above also includes the objects without WISE counterparts due to increasing performance in relation to having two separate models for objects with and without WISE counterpart
- Excluded the model_flag column (due to the item above)
- Replaced ALLWISE magnitudes with unWISE magnitudes
- [GAIA model] Trained new classification model that includes GAIA parameters. We included the results in separate columns:
- CLASS_GAIA, PROB_STAR_GAIA, PROB_QSO_GAIA, PROB_GAL_GAIA
- Included the columns from unWISE and GAIA that were used for the classification
- W1_MAG, W2_MAG, Gmag, Plx, E(BP/RP), PM
More improvements are expected for the next months and a full documentation with code access (analyses are stored in another GitHub repository) will be prepared.
Name | Description |
---|---|
ID | S-PLUS ID |
RA | Right ascension in degrees |
DEC | Declination in degrees |
W1_MAG | unWISE W1 magnitude in Vega |
W2_MAG | unWISE W2 magnitude in Vega |
Gmag | GAIA G-band mean magnitude |
Plx | GAIA parallax |
E(BP/RP) | GAIA BP/RP excess factor |
PM | GAIA proper motion |
CLASS | Class (0= QSO, 1=STAR, 2=GALAXY) from [Base model] |
PROB_STAR | Probability [0,1] of a source being a star from [Base model] |
PROB_QSO | Probability [0,1] of a source being a quasar from [Base model] |
PROB_GAL | Probability [0,1] of a source being a galaxy from [Base model] |
CLASS_GAIA | Class (0= QSO, 1=STAR, 2=GALAXY) from [GAIA model] |
PROB_STAR_GAIA | Probability [0,1] of a source being a star from [GAIA model] |
PROB_QSO_GAIA | Probability [0,1] of a source being a quasar from [GAIA model] |
PROB_GAL_GAIA | Probability [0,1] of a source being a galaxy from [GAIA model] |
Metric | QSO | STAR | GALAXY |
---|---|---|---|
Precision | 0.930 | 0.984 | 0.972 |
Recall | 0.930 | 0.975 | 0.979 |
F1 | 0.930 | 0.979 | 0.975 |
F1_weighted = 0.971
Metric | QSO | STAR | GALAXY |
---|---|---|---|
Precision | 0.939 | 0.989 | 0.981 |
Recall | 0.940 | 0.984 | 0.984 |
F1 | 0.939 | 0.986 | 0.983 |
F1_weighted = 0.978
With the environment.yml file, you can create a conda environment with all the packages needed to run the scripts.
```
conda env create --name sqg_dr5 -f environment.yml
```
```
conda activate sqg_dr5
```
```
python ./sqgtool.py --help
```
```
python ./sqgtool.py --input_folder /storage/splus/Catalogues/iDR5/VAC_features/20231121 --output_folder /storage/splus/Catalogues/VACs/sqg/iDR5 --crossmatch --n_threads 8 --verbose
```
Error logs for the crossmatch and classification processes are stored in /logs/ folders
- Please cite the original publication (S-PLUS DR2): https://ui.adsabs.harvard.edu/abs/2021MNRAS.507.5847N/abstract
- Publication for S-PLUS DR5 is in preparation.
This version of the star/quasar/galaxy classification was done in collaboration with (in alphabetical order):
- Gabriel Fabiano de Souza (IAG-USP)
- Gabriel Jacob Perin (IME-USP)
- Pierre Ré (IAG-USP)
- Raquel Valença (IAG-USP)
- Vitor Cernic (IAG-USP)
We thank Elismar Losch for providing help with data, and Gustavo Schwarz for helping with codes and server issues
We stress that the [GAIA model] classification is still under tests by our team. Changes/improvements might be expected for this model in the near future. If you find any problem, please open an issue in this repository or contact lilianne.nakazono at gmail dot com