To Grind or to Automate?
During the Barataria work, mapknitters have found that sorting the images from flights to be a time limiting step. Many of us work in the windows environment, and use windows or (free) picasa preview to quickly scroll through thousands of images, but selection can still take hours.
There's a general guideline that a map shouldn't be made from more than 30 images, or you should consider how to map the area at a higher altitude. Experienced cartographers can handle more than that, but it seems that most people will probably only have patience for ten images or so. Alex Stoicof, as an experienced mapmaker, expressed a bit of frustration that she could georectify imagery in QGIS much faster than stitching in mapknitter, and would have mapknitters use that open source method.
There has been repeated discussion of using Visual SFM to stitch images, and thus let the computer decide which images are the most useful. This has not yet been actualized, as Chad Netto is the only one which experience using Visual SFM to do such a computationally intensive task.
Numbers, Names, and 'A,B,C's'
get rid of water, shoe shots, find the b's and c's
Initial screening of the ~2000 photos on an SD card can include notes on the ranges of numbers containing shots of the launch and the landing --photos that might be useful for "aeries" or "selfies" or expository imagery (C), but not for mapping.
Similarly, photos series often contain oblique shots that have no mapping value, but can be beautiful illustrations of the landscape nonetheless (B).
It generally saves time to skip to the highest altitude shots first, just like when mapknitting. This can give you a better mental representation of the features of the landscape as you make your photo selection. I find that I have to make a mental map of how the features fit together to make a decent photo selection. This takes more brainpower than you would think, especially for tricky wetland photography.
The number of the filename allows a mapper to locate a similar photo in the series, if that is ever needed again, so the file number is information worth keeping.
Naming the file makes it searchable by string.
In windows file explorer, images can be viewed as thumbnails (see photo), and the thumbnails can be viewed at different resolutions by selecting such in the "View" menu or using the scroll wheel to enlarge the images.
a little history
During initial meetings of mapknitters in 2013, Scott showed his method, which entailed re-naming the images to include descriptions.
--changing filenames from "IMG1234" to "IMG1234 west wetland berm unplanted"
with the idea being that images would then appear as distinct, even when only viewing the filenames, enabling an informal selection that was simple to drag and drop to google drive for collective stitching effort in mapknitter.
Jenna DeBoisblanc had the idea of re-naming the best mapping photos with alphabetical prefixes,
"IMG1234" to "a IMG1234."
This would promote good map photos to the top of the filenames in a detail view.
Scott had the idea, then, of "b" and "c" photos --"b" meaning useful obliques, and "c" meaning "aeries" or funny shots with other uses.
Thus, from ~2000 images, the sort will contain an alphabetical list of photos, sorting on their utility and retaining the image number, in case mappers want to return to the original imageset for a similar image geographically close to a previously selected image.
Thus, a collection of best images might end up like this:
a IMG1234 start map
a IMG1276 pond feature
a IMG1345 boat alexis aerie
b IMG1187 cool oblique surge barrier feature
c IMG1445 alexis
For Infragram, numbers can be correllated by name, i.e.
a IMG1234 RGB IMG1344 IR
so that their matches can be easily found
From here, the numbered photos would be uploaded to google drive for collective use. Placing them in the top of the list makes it simpler to drag and drop photos.
We should be recording how much time it takes for trained and untrained, experienced and inexperienced mapknitters to sort images, so that any software authors can think about what processes are worthy of automation and which processes could be programmed with human instruction.