Figures for paper

[ismael-mendoza]

General comment: Would be great if all of these figures can be produced out of the box with functions in BTK. Consider a single notebook that creates all of these figures (except galcheat table ofc).

• [ ] In the galaxy models section: output image from BTK, 3 blends and show blended scene and individual galaxies, true locations and maybe segmentation. It would be nice to show show side-by-side parametric and COSMOS blends (@ismael-mendoza).
• [ ] table in galcheat section showing all the information provided by galcheat or an example galcheat of command line output (@mpaillassa)
• [ ] Example images for each metric (@thuiop, @ismael-mendoza):
  • [ ] Efficiency matrices comparing SEP and basic peak finding.
  • [ ] Detection metrics: Precision/recall/f1 as a function of snr/magnitude/blendedness for SEP and basic peak finding
  • [ ] Segmentation metrics: True segmentation comparison with Scarlet/SEP and IOU score in plots. Or IOU binned by blendedness/SNR
  • [ ] Reconstruction metrics: Histogram PSNR, MSR, SSIM comparing Scarlet/SEP or median binned by blendedness/SNR.
• [ ] Flux photometry measurement comparison between Scarlet and SEP (@aguinot , @ismael-mendoza)
  • Perhaps residual scatter plot of each method, where the residual is taken w.r.t to isolated noisy image flux measurement. The deblender one is taken by subtracting each of the other debledned images from the true noisy blend and using aperture photometry on that. Use true locations for all measurements.
• [ ] Maxime's HSC validation figure, is there a way to include it in this same notebook?