Uploading images with bounding boxes for object detection

Help
#1

Question/Issue:
I have already generated a set of images with bounding boxes in Open Images CSV format, so each directory has a file named _annotations.csv with following content:

ImageID,LabelName,Confidence,XMin,XMax,YMin,YMax
image_0.png,yellow,1,0.440625,0.5359375,0.5197916666666667,0.6489583333333333
image_1.png,yellow,1,0.25078125,0.3421875,0.41354166666666664,0.53125
...

I have also placed a file class-descriptions.csv in the main folder with one line only (I have only one label):

yellow,yellow

I tried to upload this folder using “Upload data” button in Data acquisition tab, there were no errors detected, I can see all of my images, but no labels and boxed are there. I also had a window with " Open Images CSV" selected, so I assume the format was recognized correctly.

What am I doing wrong?

Project ID:
516227

Environment:

  • Platform: PC with Windows 10
#2

Hi @polarisru

Have you tried to do this via the Python SDK, or what method was tried? If its the Python SDK there is an example here that you can follow

After uploading, you should be able to see your images and bounding boxes in the Data acquisition tab of your project.

Best

Eoin