What is the entire URL out of the browser, exactly where the error occurred?
Can you describe what the issue is you’re experiencing?
there was an error with task 774f25cd-9289-411b-8418-1fe3800b3039: No data to match against, please refine your matching parameters.
When I hover on the image I see A class:grouper_nassau_bicolor
This is not something that was assigned during the bulk upload metadata. Is this class assigned by the platform? Is the issue that there are no other images that have been classes as bicolor yet?
Can you provide steps on how to reproduce what you’re experiencing?
Go to encounter and click on match results
If this is a bulk import report, send the spreadsheet to services@wildme.org with the email subject line matching your bug report
This result happens generally when a matching set is specified against criteria for which there is nothing else to match against.
This is not something that was assigned during the bulk upload metadata. Is this class assigned by the platform?
Yes, machine learning applied this label. grouper_nassau_bicolor is not currently configured for computer vision matching in our configuration file. Only grouper_nassau as a machine learning label should proceed to matching
I see in previous project emails a request to map grouper_nassau_bicolor to grouper_nassau, but I don’t know if subsequent conversations changed this.
Can you please confirm that grouper_nassau_bicolor should be mapped to grouper_nassau and be matchable under that class? If so, I can fix the configuration and update the data.
These issues above should now be resolved. All 13 grouper_nassau_bicolor annotations in the database are now matchable. You’ll have to re-run matching to see these new results. The bicolors will be saved as and matched against grouper_nassau for future submissions as well.
Ok thanks. I think i misunderstood the process. i thought the bicolor would still be saves as grouper_nassau_bicolor but matched against both bicolor and grouper_nassau. Is that not possible.
If that is not possible is there a benefit to leaving them classed at grouper_nassau_bicolor or is adding a bicolor keyword sufficient?
The annotation type (e.g., “grouper_nassau” versus “grouper_nassau_bicolor”) is a strict mechanism we use to ensure we’re not trying to match unlike things, such as different species or different body features (e.g., fin versus fluke).
However, we can intercept the grouper_nassau_bicolor type and label the photo with a keyword of your choice, such as “Bicolor”, before changing it to grouper_nassau if you would like. This would provide the searchability you might be looking for between bicolor and non-bicolor while allowing bicolors and non-bicolors to be matched.
Hi @jason
If that is an easy thing for you to do that would be great. We would otherwise enter the keyword manual as we checked the encounters.
Thank you,
Alli
This should be implemented now. When computer vision detects grouper_nassau_bicolor, it should record a “bicolor” keyword on the image before saving the detection as grouper_nassau for comparability. This applies to new submissions moving forward. Do you have a bicolor image convenient to test this with?
So if I’m understanding right, we just need to upload images of bicolor nassau grouper, and they should automatically be labeled with the “bicolor” keyword on the encounter?
OK, I dug into the code, and it looks like the code to add the keyword happens after we change the grouper_nassau_bicolor class to grouper_nassau, so that setting to save the keyword would only happen if we keep the grouper_nassau_bicolor as it was originally. I do think that original solution is worth considering as bicolor is a great visual differentiator that can filter out false positives in matching.
For now though, bicolor is being matched to the non-bicolor fish. The bicolor keyword will need to be manually applied or set in the Encounter.mediaAsset0 (or 1,2,3, etc.) column in the bulk import.
Thanks @jason We can just add the keyword manually of through the bulk upload.
I will talk to Brice about changing it back to the bicolor class and the benefits of this. We would like to have individuals with both the bicolor image and the other color phases associated with them. is this possible if we keep them in separate classes?