Zoom Logo

Camera Trap Computer Vision - Shared screen with speaker view
Siyu
02:16:11
Could you please put the link to this presentation in the Google Doc as well?
Sara
02:16:57
Seconded. Also, my computer is rapidly losing battery, so please forgive me if I disappear.
Hayder
02:18:10
@Siyu, sure!
Lucy F Fortson
02:23:22
I’m going to have to dive off! This was really really great and we should find some way to keep this group going. Thank you so much for organizing Ariel.
Dan Morris
02:24:16
I mentioned that we've started posting recommended train/val splits based on location, to (a) encourage people to split their data this way for published results, and (b) allow common benchmarking. Maybe most relevant is Snapshot Serengeti, for which we've posted recommended splits here: http://lila.science/datasets/snapshot-serengeti
Jason Parham
02:24:34
Horse - Equus ferus caballus
Sarah Huebner
02:34:52
I'm sorry, but I need to leave as well. Please feel free to email me if you have questions about the specific ways we're combining ML and citizen science.
Sara
02:35:58
Hi everyone!
Sara
02:36:10
Sorry, my laptop died
Sara
02:36:20
And apprently my desk machine won't share
Sue Townsend
02:37:32
what is the competition? can you expain?
Sara
02:38:02
The cool thing is that we're getting a lot of traction in the computer vision community, which is a very good thing for all of us
Sara
02:38:29
And the other thing to note is that we are explicitly not using top-1 accuracy
Sue Townsend
02:38:39
does that mean that there is 326 different models out there?
Sara
02:38:43
Yep!
Sue Townsend
02:39:22
any way to capture all the know how...
Sue Townsend
02:39:26
?
Sara
02:39:56
And the competition metric is macro-averaged F1 score, which places more weight on recall and rare classes
mikey tabak
02:40:08
Thanks everyone for a great meeting. I have to step out, but look forward to learning more from y’all in the future.
Dan Morris
02:40:22
https://www.leonardodicaprio.org/resolve-trailguard-ground-sensors-for-advanced-conservation-monitoring/
Sara
02:40:44
We will invite the top teams to share their models and metrics for a cash prize, and will be discussing at the FGVC workshop at CVPR in two weeks
Sue Townsend
02:41:24
FGVC? CVPR?
Sara
02:43:27
Fine grained visual categorization
Jason Parham
02:43:31
CVPR - Conference on Computer Vision for Pattern Recognition
Sara
02:43:36
computer vision and pattern recognition
Sara
02:43:57
Thanks everyone!
Jason Parham
02:43:58
CVPR is about to happen in Los Angeles in two weeks, one of the premier computer vision conferences