Deblended footprints for DC2 data

Support
, , , ,
#1

Hello everyone! I am new to using the science pipelines software and the data products, and was hoping to get a little insight into what DC2 data products are available and if not, how to create them so I can start doing some science. Specifically, I would like to get the deblended segmentation maps/footprints of individual detected objects in the DC2 public data release coadds. To my understanding, this information would be in the deepCoadd_deblendedFlux catalogs, using footprint.getSpans(). It doesn’t seem that the deepCoadd_deblendedFlux intermediate data product is publicly available, so I was wondering the best way to access or create this data.

I am a DP0 delegate and I’ve looked into the deepCoadd_deblendedFlux catalogs available on the RSP, and this seems to be exactly what I need. However, I wanted to use the public data because I’ll need to have it on my own machine for some ML training and testing using GPU resources, and I’d like to make a public repo for future use. It seems like I may need to run the detection/deblending tasks on the coadds, but I just wanted to see what my options were.

#2

Hi Grant,

Thanks for reaching out. As I think you already know (you mention being a DP0 delegate) the DESC’s DC2 has been reprocessed with the LSST Science Pipelines and the resulting data products are currently being served as Rubin Observatory’s Data Preview 0.2 (DP0.2). This does include the results of deblending, including seg maps and footprints. You should not have to rerun detection or deblending tasks on the deepCoadds.

If you haven’t already, maybe work through DP0.2 tutorial notebook #10, which covers the data products of deblending that are available to users? Let me know if that’s helpful?

#3

Hi Melissa,

Thanks for the reply! The notebook will be very helpful. My main concern was the infrastructure of the RSP in terms of what I would like to use the data for - processing a large amount of data and transferring it for training. If it ends up being unfeasible, I am getting some support from DESC for looking at the DC2 data on NERSC.

1 Like