search

rfcx / arbimon

Ecoacoustic analysis platform empowering conservationists to analyze acoustic data and to derive insights about the ecosystem at scale
https://arbimon.org
Apache License 2.0
0 stars 1 forks source link

CNN validation - summary numbers don't reflect with the filters #2000

Closed koonchaya closed 1 month ago

koonchaya commented 1 month ago

Context

  1. Clear grouping: the total number of detections is not showing the current number from the selected filter. It seems to keep the previous number from the last filter and it will change to correct number when reload the page.

https://github.com/rfcx/arbimon/assets/87486094/363d84ba-0bb5-4a31-8634-6f6072bc8e8f

  1. Minimum confidence: doesn't show correct result. When I input minimum confidence number 0.1 the result is now showing only 0.1 but when I checked on best per site/best per site per day, I can see the score >0.9. This could be related to this issue https://github.com/rfcx/arbimon/issues/1999.

https://github.com/rfcx/arbimon/assets/87486094/87ce373c-e469-4905-8d15-9da9fc93e272

naluinui commented 1 month ago

Fixed in https://github.com/rfcx/arbimon/commit/e88991b68cac6081b0d949ad8fbe2266d6ab12e0 https://github.com/rfcx/arbimon/commit/b4e1b8097a2ff5e961f91a0297094ce49957824f https://github.com/rfcx/arbimon/commit/854e840c1e63efb94961ce229cadb5564626a192

naluinui commented 1 month ago

@koonchaya I think the test case you provided for point 2 was related to the "best per..." api which already described in #1999. Have you got anymore test case to prove the bug about min confidence?

koonchaya commented 1 month ago

I don't have anymore case about minimum confidence.

naluinui commented 1 month ago

FYI @koonchaya this is now on staging

grindarius commented 1 month ago

Released in v1.2.3