I have a collection of many thousands of underwater images of fishes (and other organisms to be dealt with later). As iNaturalist is heavily biased by what people image with their phones (plants, bugs) I thought it might be useful to contribute these images tagged with the proper species name to the project. In some cases these images were taken on research dives and I’m going through the dive logs to match dates/times to the photos so I can fill in GPS coordinates for the dive site for precision location information in the metadata.
I have many other images taken on dive vacations where the images were all within a short distance of a dive resort (e.g. Tawali Dive Resort, Milne Bay, Papua New Guinea). There is no way at this point of assigning precise GPS coordinates where the boat was located on these dives but they all would be within a few miles along the coast from a known GPS location.
I have a researcher friend who said he would help me with the Python script bulk upload of my images once I have finished filling in the location information in the image metadata. The images that I have exact GPS coordinates for I believe will import without issues. I’m curious how to go about doing this for localities where I only have approximate location information. Any tips or guidance will be appreciated.
Cheers.
KM
Welcome to the iNat Forum, KM!
These threads might give you some tips and guidance:
Wonderful! I knew I couldn’t be the first person to deal with this. Thanks.
KM
You can set the coordinates to the location of the resort and then set the uncertainty/precision/accuracy to the number of miles you think would encompass all of the locations. If you want to do that through the image metadata rather than the web uploader, according to this thread the EXIF tag is GPSHPositioningError.
I’ve never tried uploading this way but my one concern is that I don’t know if it would be possible to combine multiple photos into one observation through this method? There’s often value in having multiple views of the same organism whether for identification or ecological information.
I think the big thing to do is just make certain that the true location is within the accuracy circle. Put the center of the circle at your best guess for position, and then set the accuracy value large enough that it encompasses any location you think you might have potentially been.
That is a point I have not considered. The vast majority of my images are of different individuals but there are times where I have multiple images of the same individual (a fish) as it swims around and is photographed in different orientations and views. I figure the diversity of views of the same individual would be good fodder for the computer vision AI model and would be of use to upload. I will have many thousands of images of a diversity of marine fishes from around the world and will need to automate some of this in order to get it done but I may need to consider pulling out blocks of imagery where there are a multitude of images of an uncommon species sighting.
That should not be a problem. On research dives where we ventured far from the research vessel during the day in the smaller boat to reach specific dive sites I have exact coordinates where the dive boat was positioned. Those dives represent sightings within roughly 50 m radius of the boat. For recreational (vacation) dive locations, the dive resort’s dive boats generally had their sites within 1-2 miles from the resort to keep the transit distances short. Diving in areas of relatively uniform reef structure means that there is no reason for long boat rides.