A few months ago, the Wildlife Insights team conducted a survey to learn more about camera trap data management practices, the use of AI for camera trap image review, and barriers to sharing camera trap data. We promised to share a summary of the results with the community, so… here we are.
First, though, a caveat: we circulated this survey through our social networks, through conservation technology discussion platforms (e.g. see our original post to WILDLABS), and through platform user groups (e.g. through the Wildlife Insights user group). We knew that some of these starting points would bias the results toward early adopters of cloud and AI tools, so we tried our best to “break out” of our network via word of mouth circulation. But the results indicate a level of AI adoption that is inconsistent with anecdotal experience: 82 of 126 respondents indicated that they either use AI or use an AI-enabled platform. We’d like that AI adoption rate to be true, of course, but it’s orders of magnitude off of what we experience when we get unbiased samples of the camera trap community (e.g. when we're at ecology conferences). This strongly suggests that we failed to “break out” of our network, which is why we aren’t trying to formally publish these results.
And of course, we were able to send this survey to the Wildlife Insights user list, but did not have access to user lists for other platforms, so the reported use of specific platforms is likely almost entirely a function of the distribution channels, and says little or nothing about the actual user bases.
So, take these responses with a grain of salt!
Nonetheless, we believe that these responses contain some interesting data about the sectors within which camera trap researchers work, the likelihood of data dissemination after a camera trap survey, and - most interesting to us - barriers to data sharing, and barriers to the adoption of cloud-based tools.
We are summarizing only the structured questions here (i.e., all the “check all that apply” questions). Each of those was accompanied by a free-text request for more detail; anecdotally, free-text comments on those questions did not carry information above and beyond what’s visible in the structured responses. The only free-text questions represented here are the last two questions, about which image management and/or AI tools respondents are using. For those, we processed the text to make reasonable corrections (e.g. to handle spelling inconsistencies), and for consistency of interpretation, if someone named an AI model that’s not a "platform" per se in their response to the “which platform?” question, we counted that toward the “which AI tool?” question, and vice-versa. Systems that could reasonably be thought of as both a platform and an AI tool (e.g. Agouti, TrapTagger, Wildlife Insights) were counted toward both questions.
YMMV, but hopefully this starts some interesting discussions... feel free to chime in on this thread with your own answers to the original questions!
--------
Total number of responses
126
What type of organization do you work for? Check all that apply.
125 responses
Academic: 74
Public sector: 36
Non-profit: 27
Commercial: 6
Public benefit company / B-corp: 1
Other: 6
Where do you primarily conduct camera trap surveys? Check all that apply.
126 responses
North America: 66
Africa: 21
Europe: 17
South America / Central America: 14
Asia: 12
Australia: 8
Island regions: 4
Why does your organization collect camera trap data?
126 responses
Camera trap data supports my research: 110
My organization has a jurisdictional mandate or policy to monitor species (please elaborate below): 33
Camera trap data collection is among the services we provide to clients: 13
Other: 14
About how many camera trap images does your organization collect per year?
126 responses
>5,000,000: 7
1,000,000-5,000,000: 21
100,000-1,000,000: 49
20,000-100,000: 28
1-20,000: 12
I am unsure, as the data are distributed: 9
What tools do you use for reviewing your camera trap data? Check all that apply.
124 responses
I use a desktop tool like Timelapse or Reconyx MapView that's specifically for camera traps: 70
I use a cloud-based platform (e.g. Wildlife Insights or TrapTagger): 39
I move images to folders named by species: 26
I review images in an image viewer, and enter results into a separate document, e.g. an Excel spreadsheet: 26
I use a tool like Adobe Bridge or digiKam that's for general-purpose image management: 14
Other: 18
[see a summary of specific platforms from free-text responses at the end of this post]
If you're not currently using a cloud-based platform for image management, why not? Check all that apply.
108 responses
Local (i.e., non-cloud-based) tools are working fine for me, I don't have a need for a cloud-based platform: 47
Network connectivity issues prevent me from using a cloud-based platform: 27
Institutional policy prevents me from using a cloud-based platform: 27
I'd like to use a cloud-based platform, but it's a lot of work to switch from my current workflow: 19
I've tried one or more cloud-based platforms, and I didn't have a good experience: 13
Other: 24
I'm already using a cloud-based platform: 23
[see a summary of specific platforms from free-text responses at the end of this post]
Do you currently use any AI tools to accelerate your data review?
126 responses
Yes (please specify which tools, and comment on your experience, below): 83
No (if you've considered AI but haven't used it, please specify why): 43
[see a summary of specific tools from free-text responsesat the end of this thread]
Do you create geospatial analytics products from your camera trap data? Check all that apply.
125 responses
Distribution maps: 77
Occupancy maps: 73
Relative abundance: 69
Species richness maps: 52
No, I don’t map my camera trap data or use geospatial data: 25
Other (please specify): 9
After you finish analyzing your data, what are the primary means by which you disseminate your high-level analysis results (e.g. occupancy maps, covariate analyses, etc.)? This question is not about raw observational data or images, which will be discussed in subsequent questions. Check all that apply.
126 responses
Journal papers: 102
Public reports (other than academic journal papers): 68
Supplementary geospatial data (e.g. shapefiles indicating species range maps): 30
My organization does not release results outside of the organization: 12
Other: 23
After you finish analyzing your data, how do you disseminate your raw observational data (lat/lon/time for each observation)? If you disseminate most of your observations, but don’t release, e.g., sensitive species locations or human activity, that’s OK, please answer this question for the data you do release. This question is not about images, just the locations and times of species observations.
115 responses
We release data on a platform specifically for camera trap data (e.g. Wildlife Insights): 29
We release data on a platform for general-purpose data release (e.g. Zenodo): 16
We release on a platform specifically for biodiversity observations (e.g. GBIF): 14
We don’t disseminate raw observational data: 74
If you release some of your raw observational data (locations and times of species observations) to a platform for data sharing, what is your motivation for doing so? Check all that apply. This question is not about images, just the locations and times of species observations.
66 responses
Our group believes that data release benefits conservation: 46
Publication venues require it: 22
My funder (if different from your organization) requires it: 9
My organization requires it: 8
Publication venues don’t require it, but it helps with acceptance: 7
Other: 11
If you don’t release your raw observational data (locations and times of species observations) externally, why not? Check all that apply. This question is not about images, just the locations and times of species observations.
85 responses
I am concerned about improper use of fine-grained data, for example the locations of sensitive species: 45
Organizational policy prevents this: 22
I don't have time: 16
Law / governmental policy prevents this: 11
Other: 26
Do you publicly disseminate your camera trap images? If not, why not? If you disseminate most of your images, e.g. you release all images not containing humans or sensitive species, that's OK; please answer this question for the images you do release.
115 responses
Yes, I publicly disseminate my camera trap images: 33
No, I don’t disseminate camera trap images because I’m unsure where to host such a large volume of data: 38
No, I don’t disseminate camera trap images because legal or institutional policy prevents this: 32
No, I don’t disseminate camera trap images because the images themselves don’t have enough value to conservation to justify the extra work: 15
Other: 21
Do you use any other data sources to monitor biodiversity at your sites? Check all that apply.
123 responses
Acoustic sensors: 56
Drone / aerial imagery: 39
eDNA: 20
No, only camera traps: 40
Other (please specify): 32
How did you hear about this survey? Check all that apply.
126 responses
Direct email / word of mouth: 75
Twitter: 21
Wildlife Insights discussion group: 13
Timelapse user list: 12
WILDLABS: 7
LinkedIn: 6
AI for Conservation Slack workspace: 1
Other: 4
Which platforms/desktop tools do you use to process camera trap images?
Timelapse: 35
Wildlife Insights: 23
DigiKam: 9
Custom/in-house: 7
Zooniverse: 5
EcoAssist: 4
camtrapR: 4
Excel: 4
Camelot: 3
Trapper: 2
Agouti: 1
Wild ID: 1
Photo Mechanic: 1
Camera Base: 1
Animl (R package): 1
Animl (online platform): 1
TrapTagger: 1
CPW Photo Warehouse: 1
IrfanView: 1
Camera Trap File Manager: 1
Zamba Cloud: 1
Which AI tools do you use to process camera trap images?
MegaDetector: 61
Wildlife Insights: 23
Custom/in-house: 12
EcoAssist: 5
MegaClassifier: 2
Reconn.ai: 1
DeepFaune: 1
Agouti: 1
Camelot: 1
Data Camp Workspace AI: 1
TrapTagger: 1
Zamba Cloud: 1