saulgreenberg
commented
5 years ago Another user said the following:
- Data format/template: For each animal we detect, we need to know a series of things about the animal, some are standard (species, age, sex) and a few items may be project or species specific (antlers, markings present etc).
Not a problem if there is only 1 animal per photo or all animals have the same attributes. But, sometimes there are 10 elk, including a mix of adult and juveniles, some close and some far. Or a magpie (near, sex/age unknown) and a deer (near, female, adult). In the past, people have made templates that are very detailed, with a counter for all important combinations: Elk_Male_near, elk_female_near, elk_young_near, elk_unknown_near, elk_far --- and so on, then comments section for rare species.
In my mind, this works best if you use the click-marker, then for each click, you get a drop down with your 3-5 animal specific questions (spp, age, sex, project_specific_Attribute1, etc.). You can sort of “add individual” in the data entry area and get the same options for a new animal. It would essentially allow >=1 rows for each photo in the database. I feel like this could almost be made into a default template and then the flexible option would be to go to the template. This would potentially help people new to the program/concepts to use a standard default set up, rather than reinvent a database from scratch.
I think some changes in how we use and see large data sets, probably in the last 5 years, has generated a push toward “tidy data”, at least within the community of geeks that I participate (R/data analysis).
For cam trap data for me, tidy data would mean each “animal type” within a picture would be a row. If there are 3 close female elk, that can be 1 row or 3 rows (depending on how the person prefers to count; it’s easy to collapse tidy/standard data). But if there is also a turkey, or 1 male elk in the photo, that is a new row (repeating the same photo id).
By moving in this direction, most of the other columns would be ~standard, making it easier to combine datasets for meta-analysis and giving me a more uniform starting point for subsequent code. Can always allow additional columns that are default NAs (antlers, e.g.) I think the key is just allowing more than 1 row per photo.
In my mind, the standard column names would be Picture data, species, sex, age, count, behavior, proximity, project_or_species-specific_column1,2,3
I think the user could specify all of animal type 1, then click on something that allows you to also identify animal type 2 [“move to next picture” or “add animal”]. Snapshot Serengeti has a good interface for this. You can keep adding types of animals, then click done. Some things might not need to be “individual specific” but could be applied to all individuals of a species. I like how snapshot Serengeti does this with “young present”. Anything species-specific could be set to default to NA (like antlers).
A key bonus of timelapse for us, is the ability to go quickly through photos with nothing in them. So, keeping the ability to set up a “0” and copy it forward easily would be important.
I think a conditional fields feature would be useful, e.g. we only need age/sex for some species, only need boat details if a boat is in view. I think we are talking about changing what the user sees (or is instructed to do) while at the same time each row in the database will include all fields, and conditional fields remain blank or NA when the condition is not met. The boat example would be obvious for the user, and for simple conditions I find it works fine to say, in hover-over text, 'only required for XXX', plus written/verbal advice to first-time assistants. For complex conditions, perhaps you could allow users to set up 'sub-lists' of pick-list items for which the condition should apply. Then set up corresponding field-lists to specify which extra fields will be required. Maybe condition could potentially be a complex if/and/or statement? Not sure how this would be implemented in the interface, perhaps hide/show extra fields, or perhaps open a subsidiary window with the extra fields? For me this sounds a nice wish-list item but not a high priority for current work.