Skip to main content

Zoom Logo

WILDLABS Virtual Meetup: Big Data in Conservation - Shared screen with speaker view
Stephanie O'Donnell
38:19
@Thomas - here you go! https://www.nationalgeographic.org/labs/exploration-technology/
Anton Alvarez
38:46
Hi everyone: Anton Alvarez — collaborating with WWF Spain — Camera traps
Jon Van Oast
41:46
hi anton - we work with wwf spain on iberian lynx. :)
Michael Evans
42:42
Hi Everyone, Mike Evans with the Center for Conservation Innovation at Defenders of Widlife - work on remote sensing and machine learning
Dan Morris
49:40
[Virtual applause]
Rob Appleby
49:44
Thanks Dave!!!
James Arkell
49:58
Its strange not hearing any applause!
Iding Achmad Haidir
50:05
Thank!
Kate Wing
50:15
Qs!!
Rob Appleby
50:16
is that DNA down to the individual from paw prints?
Kate Wing
50:25
Can you talk about how you pick vendors and specs?
Kate Wing
51:10
Are the slides going to be posted at Wildlabs.net?
Stephanie O'Donnell
51:37
@Kate - what do you mean by vendors and specs, in what context?
Stephanie O'Donnell
51:44
W’e
Talia.Speaker-intern@wwfus.org
52:03
We’ll post the recording on WILDLABS so you can look back at them
Kate Wing
52:40
I wasn’t able to see Dave’s slide so when he talked about the Big Data Ecosystem I couldn’t see if he had broken down different products/vendors by types of services. Many orgs don’t know whether they should be using AWS or Azure or Splunk or… How did Dave T decide how to build his stack?
Aurelie Shapiro
53:04
there were all these vendors by type it was a cool slide
Stephanie O'Donnell
53:07
It’s such a good question, one I’ve been wondering about as well.
Stephanie O'Donnell
53:12
We can revisit in the discussion.
Robert Elliott
55:40
I’d like to hear about shared data standards and formats. Thanks
Thomas Starnes
56:07
^ Upvote common standards
Stephanie O'Donnell
56:10
That’s a big discussion point that came through in the questions we asked in registration
Robert Elliott
56:26
👌🏽
Kate Wing
56:37
Also +1 common standards
Thalie Partridge
56:54
Yes, shared and open data and managing sensitive data issues.
Jon Van Oast
57:19
same. :) (i was at the digital biodiversity data... it seems like a universal desire -- not so easy to execute. heh)
Stephanie O'Donnell
57:24
Yep. Standards and how to share/collaborate was the most common question people wanted answered/discussed.
Aurelie Shapiro
58:27
it would be cool to be able to share more data, maybe by combining it with other data or something...there's a lot of visible data in Movebank but it is not always accessible for analysis (for obvious reasons), so maybe it could be blurred or degraded somehow and still useful
Jon Van Oast
58:30
if anyone is not familiar -- might be of interest. https://www.idigbio.org/content/digital-data-biodiversity-research-conference
James Arkell
58:32
does common standards refer to consistency of formatting? If so I'd like to hear about how we can overcome formatting differences :)
Stephanie O'Donnell
58:52
We had the chair of the data logging society speak in our last meet up.
Stephanie O'Donnell
01:00:03
This is a big questions.
Rob Appleby
01:00:27
Thanks Sarah
Jon Van Oast
01:00:30
thanks!!
Jon Van Oast
01:02:29
hey dan!!
Karin Schwartz
01:03:41
Hi Dan, can you speak a little more slowly since this is an international meetup - thanks
Kate Wing
01:06:34
Is that demo public?
Kate Wing
01:06:48
OTHER PENGUIN
Aurelie Shapiro
01:07:01
top right?
Aurelie Shapiro
01:07:17
baby moose or whatever can be tough
Aurelie Shapiro
01:07:23
yesss
Stephanie O'Donnell
01:08:03
Dog fooding?
Kate Wing
01:08:13
Means trying your own software
Stephanie O'Donnell
01:08:24
ah! Thank you
Kate Wing
01:08:29
Because dogs, um, eat things they make
Shashank Srinivasan
01:09:19
haha thanks for the explanation!
Stephanie O'Donnell
01:09:22
13% donkey? :D
Jon Van Oast
01:09:31
heh. i always thought it was because if you work at the dog food factory, you should eat your own product to make sure it is acceptable. haha
Kate Wing
01:09:53
Leafsnap on steroids
Karin Schwartz
01:11:54
Would it be possible (or has it been done) to do a species classification system using footprint photos?
Talia.Speaker-intern@wwfus.org
01:13:00
In case anyone missed that: aka.ms/speciesclassification
Rob Appleby
01:13:12
what's next Dan? Reinforcement learning? GANs?
Rob Appleby
01:13:19
Also, why is it hard Dan?
Sade Moneron
01:13:54
How many images were/are needed to build the machine learning tool to identify a particular species. Hundreds, thousands?
Kate Wing
01:13:56
Maybe we need a follow up with Dan on the technical side bc I would like to bring in my AI partners for that (per Rob A’s comment)
Stephanie O'Donnell
01:14:26
Okay - let’s have some specific questions to dan at the end before we go into the general discussion with all speakers
Kate Wing
01:14:50
Sade - sort of depends on the question you are asking and what conditions you want to use the AI in, but 1500 still is good. Video is different.
Thomas Starnes
01:15:10
Dan - what's the difference between the Microsoft image repository and Wildlife Insights (not to mention all the rest)?
Stephanie O'Donnell
01:15:27
Thomas - excellent question. Also - how do they fit together.
Kate Wing
01:16:04
TNC is using Samasource to label and annotate video data now (from camera traps and from boat video) to train an AI. Has anyone else hired out their labeling?
Kate Wing
01:16:57
RECONYX ?
Anton Alvarez
01:17:53
Karin…search about conservationfit.org…“footprint identification technique (FIT) is one such approach; it identifies species, individuals, sex and age from footprints alone.”
Karin Schwartz
01:18:38
Thanks, Anton. I'm actually the Scientific Coordinator for Ex situ Partners for ConservationFIT!
Isla Duporge
01:19:32
Really interesting thanks Dan
Jon Van Oast
01:19:34
thanks dan
Stefan Istrate
01:19:58
A question for Dave Thau, but I don't have a working microphone to ask it live:Since you recently transitioned from Google to WWF, can you talk about what kind of impact you can have now in a conservation NGO, but you couldn't in a tech company. (Context: Google / Microsoft are both applying their machine learning algorithms to conservation problems, so a software engineer / data scientist could work on these either in the IT sector, or in the NGO sector.) Is WWF trying to attract machine learning practitioners directly? Can NGOs compete with tech companies on attracting talent?
Stephanie O'Donnell
01:20:26
Oh this is a great question! I’ll ask it in the general discussion, Stefan
Iding Achmad Haidir
01:20:28
thanks Dan, very good information
Dave Thau
01:22:07
I can help answer that one too.
Stephanie O'Donnell
01:22:45
I thought as much - I’ll open it up to you next
Jon Van Oast
01:24:35
agreed! the holy grail would be a meta-score across different cv. :D
Jes Lefcourt
01:25:55
Does that multiply total computation time / cost? Is a combination of calling multiple, individual services inefficient?
Kate Wing
01:27:00
That assumes open APIs
Thomas Starnes
01:27:14
Thanks Dan & Dave. At the recent ZSL workshop, the concept was discussed of a 'model factory' / 'model warehouse' with, as Jon says, a meta-score!
Rob Appleby
01:27:34
I'd like to know if/when it'll be possible to combine GPS and camera trap (and many other sources of...) data in real time to tell us about individuals, management actions and predictions for problems such as human-wildlife conflict etc.
Simon Hoyte
01:29:04
If we are to live up to the true values of citizen science, what mechanisms are being developed by tech companies (Microsoft) or NGOs (WWF) to enable and support anybody, anywhere to be able to contribute data?
Kate Wing
01:29:33
THAT IS THE QUESTION
Jon Van Oast
01:29:43
:)
Kate Wing
01:29:58
If I have 10k to do something, what should my screens be for who I choose to work with and what tech approaches?
Jes Lefcourt
01:30:06
And we all appreciate Google, Microsoft, etc's donation of resources!
Kate Wing
01:30:20
Yes, definitely Jens
Aurelie Shapiro
01:30:33
what about putting these algorithms in the cameras themselves - and avoid having to upload in the crowd and process. Can't a camera recognize a tiger/poacher/bulldozer?
Aurelie Shapiro
01:30:43
*cloud not *crowd
Isla Duporge
01:30:47
There is alot of focus on classfication in camera trap images- the rise of UAVS for willdife monitoring and satellites to monitor large bodied mammals gives us a different top-down perspective on species. Will the architecture developed for identifying species in camera trap images be easily adaptable for these kind of images?
Jon Van Oast
01:30:49
yes please, aurelie.
Kate Wing
01:30:52
Yes, can we get someone from NVIDIA here to talk about camera tech?
Stephanie O'Donnell
01:31:18
Sure! Let’s put that in the suggestion box for next year’s meet-ups.
Thomas Starnes
01:31:30
^ great question, Isla
Shashank Srinivasan
01:32:16
+1 Isla
Stephanie O'Donnell
01:32:28
Oh, hey shashank!
Stephanie O'Donnell
01:32:41
And Isla - will get you to put this to speaker.
Shashank Srinivasan
01:32:50
Hi Stephanie!
Kate Wing
01:33:05
My partners at CVision have a new Windows version of their open video classifying tool for fisheries video if anyone wants to try it on other video https://github.com/openem-team/openem
Dan Morris
01:33:29
@ Isla, re: whether camera trap models will translate to UAVs... I'm very cautious about saying that models will generalize from one sensor to another, even from camera traps to handheld cameras, for example. So I think we can learn broad lessons about data curation that generalize from camera traps to UAVs, but from a machine learning perspective, I would consider it almost a brand new problem.
Alexandra Dumitrescu
01:33:53
Hi:) i am online
Kyler Abernathy
01:34:09
NGS is collaborating with Resolve on 'camera traps' with on-board image analysis and discrimination.
Jon Van Oast
01:35:13
interesting, kyler. thanks for the info.
Aurelie Shapiro
01:35:14
there are lots of people looking at automated species classification of photos from from airplanes, drones might not be that much different
Thalie Partridge
01:38:07
I’m interested in how small NGOs and Indigenous people who collect the data can engage with big data and have greater agency its use? What are your data ethics policies and training recommendations?
David Wolfson
01:38:35
If you're interested in an approach for camera trap image ID that isn't cloud-based, we came out with an R package that is linked to an ML algorithm trained on 27 common North American species: https://github.com/mikeyEcology/MLWIC
Isla Duporge
01:39:16
Mentoring would be fantastic!
Jon Van Oast
01:39:54
perhaps wildlab could have tech "office hours".... regular time for ngos to meet up with a pool of tech folks...
Jon Van Oast
01:40:04
:D
Alexandra Dumitrescu
01:40:36
:*
Alexandra Dumitrescu
01:40:39
thank you
Kate Wing
01:42:27
THIS. very important to connect the content experts/problem defintion folks to the tool builders
Jon Van Oast
01:42:34
yes. that gap.
Kate Wing
01:42:35
Dave That is speaking my life
Kate Wing
01:42:39
Thao
Jon Van Oast
01:43:03
mine too (from the tech side of the gap, fwiw)
Aurelie Shapiro
01:43:04
but even NGOs have problems to drive decisions...that gap still exists
Maia Adar
01:43:43
What is the greater challenge right now: make better tools or use the tools to take actionable steps?
Jon Van Oast
01:44:12
i cant speak enough about getting tech folks "on the ground" as dave says, btw.
Kate Wing
01:44:17
Start with measurable steps and build tools that get you there
Aurelie Shapiro
01:44:23
using the tools to take actionable steps
Stephanie O'Donnell
01:45:10
Snaps to this
Rob Appleby
01:45:10
I don't want to be a Debbie Downer, but I am wondering why, if we have such great tools and fantastic data sources, are we failing in many conservation goals?
Aurelie Shapiro
01:45:28
yes! - our "real" ideas are very valuable to people looking to create tools and methods
Dan Morris
01:45:44
+1 to what Sarah is saying! If anyone needs to know, I'm dan@microsoft.com, and I love talking about conservation. :)
Iding Achmad Haidir
01:45:56
From my perspective from a developing country, I think the gap would be between the 'tools builders, the NGO and the GO aka policy makers'. We need to build bridge amongst these institutions.
Stefano Balbi BC3-ARIES
01:46:20
Great seminar, just a minor point to provoke discussion. AI and ML can be done without major Corporations. As Conservation practicioners do you trust big Corporations? I understand is not Oil business but you know corporate interest will always prevail.
Maia Adar
01:46:56
Ok so more specifically I see it as three parts — make tools, use the tools to extract insight about what should be done, or get the “decision makers” to actually do what the data said — what is most difficult?
Jes Lefcourt
01:47:15
Definitely the latter
Iding Achmad Haidir
01:47:22
yes, the latter
Kate Wing
01:47:39
Rob wins the Time Zone Endurance Prize
Jon Van Oast
01:47:53
heh
Paul Millhouser
01:49:55
Rob, I think part of the problem is that we as scientist have begun to master the technology to cope with big data, but we have a long way to go with visualization for a broader audience.
Aurelie Shapiro
01:49:59
getting the decision makers to do what the data said, or having them understand clearly what should be done...
Jes Lefcourt
01:50:08
People need to understand and care about the data. As we all unfortunately know, that often doesn't match political interests. I think every piece of data we can provide makes it harder to refute, though.
Shashank Srinivasan
01:50:32
@Maia we’re having some impact by framing conservation data from a business risk perspective directly to infrastructure companies in India.
Jes Lefcourt
01:50:33
+1 to Paul, too. It's about storytelling.
Jes Lefcourt
01:51:52
Great topics for next year. Storytelling in general, like NatGeo. How do we take data and drive impact?
Jes Lefcourt
01:54:06
The Movebank animations of migration patterns are fantastic!
Stephanie O'Donnell
01:54:14
they are amazing!
Rob Appleby
01:55:50
Dan still didn't tell us all why it's so hard!?!?!?!?!
Kyler Abernathy
01:58:17
The need to find a good way to communicate big data to the public and to decision makers is definitely key to enabling or instigating action. NGS is building on our long history of storytelling to develop new ways to visualize and share conservation data. https://www.nationalgeographic.org/labs/geographic-visualization/
Jon Van Oast
01:58:45
wondering if there is anyone from gbif on here? (re: sharing data) ... they seem to have a ton of experience in this area
Jes Lefcourt
01:59:28
So hard, understandably, for folks to be willing to set aside the fact that their data is often their primary differentiator.
Thomas Starnes
02:00:55
@Jes - so true, especially for smaller NGOs, in my experience
Unknown Speaker
02:01:16 Rob Appleby:Thanks so much to the speakers! Great stuff and very grateful!
Dan Morris
02:01:20
Thanks everyone, great to meet all of you!
Isla Duporge
02:01:22
Thanks for organising this was great!
Jes Lefcourt
02:01:25
Thank you all!!!!!
Kyler Abernathy
02:01:25
Thank you to the speakers and all who joined in!
Vance Russell
02:01:26
Thank you to all the speakers and Wildlabs for the meetup!
Alexandra Dumitrescu
02:01:27
thanks to everyone who poures their wit and passion in helping us move forward. steph please send us the recording soon to share with our teams
Cathleen Balantic
02:01:28
Thank you all for your insights and time!!
Jon Van Oast
02:01:29
will this video + slides be archived?
Rob Appleby
02:01:31
Thanks Steph and Talia!!!
Paul Millhouser
02:01:32
Thanks to all!
Dave Thau
02:01:34
Thanks all! Great to see you all!
Jon Van Oast
02:01:35
+ chat?
Anton Alvarez
02:01:37
Thanks everyone!!
Maia Adar
02:01:41
Thank yoU!!
Dan Morris
02:01:48
And thanks to Stephanie for getting everyone on this call so engaged!
Rob Appleby
02:01:49
Talia da bomb!
Dave Thau
02:01:50
Thanks Steph and Talia!!!
Shashank Srinivasan
02:01:54
Thanks a lot for organising this Steph and Talia; it was a really interesting discussion!
Karin Schwartz
02:01:54
Thanks, Stephanie and all the speakers! I'll email Dan about ConservationFIT.
Iding Achmad Haidir
02:01:57
Thank you very much to all speakers and organizers
Talia.Speaker-intern@wwfus.org
02:01:58
Thanks everyone!!
Jon Van Oast
02:01:58
so great ! -- thanks everyone
Thomas Starnes
02:01:58
Thank you! Amazing and inspiring meetup.
Afroditi Kardamaki
02:01:59
Thank you!
Thalie Partridge
02:02:02
Thanks
Talia.Speaker-intern@wwfus.org
02:02:02
More info on the series: https://www.wildlabs.net/virtual-meetups
Julia Hunter-Anderson
02:02:03
Thanks everyone, just sad I missed the other two sessions!
felipe avino
02:02:05
thanks
Rob Appleby
02:02:12
happy holidays ya'll