Table 3
Overview of wall-clock time τ, ∆ɀ, σNMAD and η for sources classified as being extended (EXT) or point-like (PSF) from a single model (as opposed to an ensemble utilized for the results depicted in Figure 7) trained with four input configurations.
| Features | τ | ⟨∆ɀ⟩EXT | ⟨∆ɀ⟩PSF | σEXT | σPSF | ηEXT | ηPSF |
|---|---|---|---|---|---|---|---|
| Model flux(a) | 11:14,41 | −0.045 | −0.043 | 0.062 | 0.096 | 11.1% | 29.1% |
| Model flux + ap. flux(b) | 16:03,82 | −0.020 | −0.012 | 0.044 | 0.058 | 5.4% | 10.2% |
| Model flux + ap. flux + Images(c) | 20:39,91 | −0.005 | −0.002 | 0.045 | 0.052 | 4.3% | 8.4% |
| Images (ɡriɀ + W 1 − W4 ap. flux)(d) | 17:17,07 | −0.029 | −0.003 | 0.043 | 0.059 | 5.2% | 9.6% |
Notes. The last two setups are an improvement over the previous performances, with the last setup providing comparatively good results, despite ignoring the features provided by catalogs, particularly relevant for potential implementation in LSST for example. (a) Catalog-based model fluxes only, as traditionally done; (b)Catalog-based aperture fluxes and respective colors as done in Saxena et al. (2024); (c) Current PICZL setup as per Table 2; (d) ɡriɀ + W 1 − W4 aperture flux images.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.