I am thinking of mapping the CTMS or the Camtrap DP to the Wildbook Bulk Import Excel. But I want to know how to proceed. I think just some fields are going to be in the Bulk Import Excel.
There are several people using camera trap data in Wildbook and I want to see how the people are organising the data.
My goal is to “automate” the processes from CTMS or Camtrap DP (or another type of data schema) to Wildbook Bulk Import Excel.
All of our called photos are labelled within Wildlife Insights and data from the export files (which I believe are the same/similar format to the CTMS) have been used for WildMe imports. We are only using a small subset of the available data fields (location, species, group size, and individual ID). Jason has kindly run some large imports for us, but it would be great if someone developed a pipeline that allowed users to easily integrate WI with WildMe bulk imports. I’m not sure if the pipeline that Jason developed (in Alteryx?) is something that users could easily use. If you pursue this, I would greatly appreciate it if you share the details.
The big issue we’ve had is dealing with multi-individual photos. For single animal photo sequences, Jason has imported these into WildMe with one trigger/photo sequence = one encounter. These are great to work with. For multi-individual sequences, photos with multiple annotations are cloned and split up into separate encounter files. If a sequence contains 20 photos with 10 deer in each photo, this can quickly balloon into an unmanageable situation. Roughly 1/2 of our 50,000 deer sequences contain multi-individual photos, so handling this issue is high on my wantlist.
Hi Ben
Yes, I think Wildlife Insights is using the CTMS too. We are also using Wildlife Insight for the Iberian Lynx, so maybe we plan to work in a pipeline between Wildlife Insight and Wildbook.
Location is Location ID? Or more fileds?
How do you know the group size? Based in the number_of_objects in the images.csv from Wildlife Insight or how?
I has thinking in one images = one encounter (our secuences are 3 or 5 images), I was worry about 2 ID in a secuence (but only one animal in each images) that I think is difficult to handle in Wildbook.
I think that Traptragger in their integration is also use a small subset of the data fields (Encounter.decimalLatitude, Encounter.decimalLongitude, Encounter.genus, Encounter.specificEpithet Encounter.submitterID, Encounter.year, Encounter.month, Encounter.day Encounter.hour, Encounter.minutes, Encounter.mediaAsset0) and one picture one encounter
On the other hand, multi-individual sequence is a difficult task to solve, I don’t know how to solve it.
I’ve seen different approaches with carnivore species in CT image sets. If you have paired CT cameras, then the approach other lynx researchers have taken is to do a multi-image encounter upload. This can be done either via the Report an Encounter functionality or via bulk import, if you have large volumes of data to get into the system.
The key with this approach is to be absolutely sure that all images going into a single encounter are of the same individual. Since lynx tend to be solitary, occasions when more than one individual is seen at the same time and place is rare so the risk of having different individuals’ photos in the same encounter is low.
For more social species, this isn’t practical so there, users will generally upload one photo per encounter - in the bulk import template, this equates to one image file name per row. This also solves for multiple individuals at the same CT event. Our user community takes advantage of the Occurrence / Sighting ID to group multiple photos together from the same CT event, which helps support multiple individuals at the same event and is the mechanism whereby Wildbook can capture co-occurrence information automatically - once 2 different individuals have been identified in the same Occurrence / Sighting, even if they’re not in the same photo together, they’ll be recorded as having co-occurred with each other.
For the single photo / single encounter approach, I know of several users have created R code to help generate their metadata spreadsheets. One of our users has generously permitted me to share this R code to others who might find it useful. It would need to be modified to support populating a spreadsheet for multi-image encounters, though.
I also have bulk import templates for both of these (single and multi-photo) scenarios as well as a Data Prep guide, for tips on other metadata to be uploaded. If you’d like copies of these and/or the R code mentioned above, I’m happy to share, just let me know.
I was thinking in created an simple windows app (GUI based in Python) for help to generate the metadata. If I have your templates, guide, tips, and R code I can try to take in acount for the app. Any sugestion is welcome!
I am somewhat concerned about having two deer in one encounter when we thought there was only one animal, but I suspect these rare cases will be swamped out with the large sample size of other images.
Yeah, I have some thoughts abut how to deal with multi-animal encounters, but I’m not sure how easy they might be to implement.
For our dataset, this quickly becomes unmanageable. How big of a dataset are you using this approach with? How many volunteers do you have working with the images?
The lower density of carnivores over herbivores generally means that we wouldn’t have the same volume of data that you have. To clarify, my organization aren’t direct users of our Wildbooks, we are admins and frontline support. Our user community is represented by various types of organizations with varying #s of resources to curate the data so I can’t really address the question re: volunteers / data curators, sorry.
Hi, I have version 0.0.1 of the app that I was doing. @ben It has the functionality to facilitate the download of images from Wildlife Insight and also facilitated the creation of the Wildbook Bulk import Excel from the Wildlife Insight data.
I have followed the advice of @ACWadmin1 to use OccurenceID to be the CT event. Also, have the option to have multi-image encounters based on a threshold time, among other functionalities.
I would appreciate some feedback.
Antón
