For HAL, OPEN THIS FILE WITH MARKDOWN READER in Jupyter Lab or Preview with Markdown Reader in HAL
This section will explain hte structure and purpose of the folders/files in this repo.
As of 11/13/24 and based on LSST/Rubin Project Outline, the most relevant notebooks and scripts are metrics and metrics_v2 notebooks and lsst_anns.py.
lsst_anns.py: Creates annotations for objs in cutouts using LSST Truth catalog info (has multiprocessing)
run_model.py: Training script.
metrics_v2.ipynb (Currently working on as of 11/13/24): Slightly improved but incomplete version of metrics.ipynb: Using a class instead to set thresholds and make DeepDISC predictions, Plots of Obj Properties, Calcualting Detection Completeness, 1-1 Catalog Matching to Truth Catalog, 2D Histogram of FOF Plots
metrics.ipynb (Currently working on as of 11/13/24): Creating DeepDISC detection catalog, LSST Truth/Detection catalog based on Test Set, Calculating Detection Completeness, 1-1 Catalog Matching to Truth Catalog, Creating 2D Histogram FOF Plots, Creating Unrecognized Blend vs Mag Plots
AddRomanLSST.ipynb: Adds Roman images to LSST images by both upsampling and padding LSST Images giving us images of dims (512,512).
galsim_truth_anns.ipynb (May not be as updated as lsst_anns.py): Notebook version of lsst_anns.py that creates annotations in multiband for obsjs in cutouts using LSST Truth Catalog
galsim_truth_anns_multiband.ipynb (Has been incorporated into lsst_anns.py): Grant's notebook adding multiband info to annotations
RunInference.ipynb: Notebook that evaluates the model and creates all the plots (AP scores, metrics, mags vs metrics, confusion matrices, etc).
unrec_blend_frac.ipynb: Notebook to calculate unrecognized blend fraction for both combined data trained model and non-combined data trained model.
AllObjsHist.ipynb: Plots histograms of distributions of Roman data.
data_processing/: Contains all the data processing scripts/notebooks used to explore/format Roman data. The subfolder lsst/ has the notebooks used to combine Roman + LSST data (starting with Roman data first and adding LSST images) and to create the LSST Detection catalog.