Open srodney opened 3 years ago
Note: in sprint 2 planning, we decided that this is low priority. We want to focus first on doing the new thing, i.e., classification into "multiply-imaged candidate" or "not that" rather than "single transient candidate" vs "not real" (which is what braai does). Resolving this issue in the future, however, will allow us to feed the braai results into the multiply-imaged classification algorithm to see if we get different / better results compared to a naive threshold-based feeder function
For example, use the braai architecture from ZTF
See the paper