Insects
Butterflies (Lepidoptera, Rhopalocera)
Nymphs (Nymphalidae)
Sighting 4468803

Oreixenica kershawi

1 Striped Xenica at Cotter River, ACT

Oreixenica kershawi at Cotter River, ACT - suppressed
Oreixenica kershawi at Cotter River, ACT - suppressed
Oreixenica kershawi at Cotter River, ACT - suppressed
Oreixenica kershawi at Cotter River, ACT - suppressed
Oreixenica kershawi at Cotter River, ACT - suppressed
Oreixenica kershawi at Cotter River, ACT - suppressed
Oreixenica kershawi at Cotter River, ACT - suppressed
Oreixenica kershawi at Cotter River, ACT - suppressed
Oreixenica kershawi at Cotter River, ACT - suppressed
Oreixenica kershawi at Cotter River, ACT - suppressed
Request use of media

Identification history

Oreixenica kershawi 6 Feb 2023 MatthewFrawley
Geitoneura klugii 6 Feb 2023 CarbonAI
Heteronympha paradelpha 6 Feb 2023 CarbonAI
Oreixenica lathoniella 6 Feb 2023 CarbonAI
Oreixenica kershawi 6 Feb 2023 DPRees125

Identify this sighting

Please Login or Register to identify this sighting.

User's notes

Very local, stuck close to damp creaklines with few venturing into drier more sunny parts narby along track

9 comments

MatthewFrawley wrote:
   6 Feb 2023
This is interesting, @CarbonAI had a few cracks at this one but missed the mark each time, I guess because it isn't trained in Striped Xenica, as it has too few photos in the database at present.
AaronClausen wrote:
   6 Feb 2023
@CarbonAI silly robot!
DPRees125 wrote:
   6 Feb 2023
yeh, it kind of got it was a brown, will be interesting to see if it can sort out the other Oreixenicas...
AaronClausen wrote:
   6 Feb 2023
@MatthewFrawley note this is with probability set to > 30% - if we set to say 70% then none of these bogus suggestions would have made the cut. We can play with that threshold over time if we think it's set too low.
AaronClausen wrote:
   6 Feb 2023
Currently deploying a small change that will:
- Display the number of learnt images on the species page
- Safeguard to prevent CarbonAI from recursively learning from it's own confirmations i.e. preventing an infinite loop where it turns into the terminator
MatthewFrawley wrote:
   6 Feb 2023
Thanks for the explanation @AaronClausen and perhaps this is a good indicator that we set the probability threshold higher than > 30% to avoid multiple suggestions that are a low probability of being correct. By the way, does the robot look at all the images in a sighting? I think if it was trained in Striped Xenica then it would have picked it with the third image here, which shows the very distinctive striped patterns on the underwings.
AaronClausen wrote:
   6 Feb 2023
You are very wise @MatthewFrawley - correct - it is currently ONLY looking at the first image. But we can easily change that. You want us to start scanning all images? I can do this in a sec.
MatthewFrawley wrote:
   6 Feb 2023
Yes, that would be good @AaronClausen - we do sometimes get sightings where the user has unwittingly taken a photo of a different species and included it in the sighting, so the robot could get confused there. The human moderator, in these instances, points it out and asks the user to delete the image and submit it as a separate sighting.
AaronClausen wrote:
   6 Feb 2023
Ok @MatthewFrawley yep makes sense, leave it with me...

Please Login or Register to comment.

Location information

Sighting information

Additional information

  • 25mm to 50mm Animal size
  • Mixed Gender
  • Alive / healthy Animal health
  • Collecting pollen Insect behaviour on flower
  • True Insect on a flower

Species information

Record quality

  • Images or audio
  • More than one media file
  • Verified by an expert moderator
  • Nearby sighting(s) of same species
  • GPS evidence of location
  • Description
  • Additional attributes
831,070 sightings of 22,884 species from 14,389 members
CCA 3.0 | privacy
NatureMapr is developed by at3am IT Pty Ltd and is proudly Australian made