Blank image classification in data download

55 views
Skip to first unread message

Lucy Hughes

unread,
Apr 15, 2025, 5:39:13 AMApr 15
to Wildlife Insights
Hi,
I have been running my project for several years and have just finished classifying season 3. When I downloaded the data I noticed an oddity about the way it showed Blank classifications on the images.csv file. In seasons 1 and 2 a blank classification has a WI Taxon code assigned, the word "Blank" in the common_name column and a"1" in the number_of_objects column. 

However, in season 3 some blanks are reported like this and others have no WI Taxon code, no name in common_name and a "0" in the number_of_objects column. here's a screenshot of the two versions. 

Screenshot 2025-04-15 112236.png

Can you tell me what is happening here. why the two different ways of reporting images classified as Blank? 

Thanks
Lucy

Melissa Kay

unread,
Apr 17, 2025, 6:30:12 PMApr 17
to Wildlife Insights
Hi Lucy,

I'm not affiliated with WI at all, just another user, but I believe I have an answer to your question. The discrepancy depends on whether the identifier hit the "Blank" button (shown below in red) or pressed, "Apply to sequence" (shown below in blue). The Apply to sequence button will give you the taxon code and common name, whereas selecting Blank will give you nothing.

Hope that helps!!

 

Melissa Kay

unread,
Apr 17, 2025, 6:31:14 PMApr 17
to Wildlife Insights
The photo didn't attach. Sorry
Screenshot 2025-04-17 162719.png

Lucy Hughes

unread,
Apr 18, 2025, 2:04:17 AMApr 18
to Wildlife Insights
Hi Melissa,
Thanks for the suggestion but this is not my issue, I don't classify  sequences, mine is an image based project. In the three years that I have been classifying the project for some images I accept  the AI suggestion of 'Blank' and for others I choose to edit the classification and label it 'Blank' Up to December 2024 all 'Blanks' on the data download had a WI taxon code, common name regardless of which method was used to classify it as 'Blank'. From December 2024 through to March 2025 I now have a mixture of some events with the taxon/common name some without. Its possible that using the two different methods accounts for this but hard to say. I have submitted a bug report so hopefully they will have some luck working it out.

Reply all
Reply to author
Forward
0 new messages