I've tried to create an ImageDB with only one image and had 75 as score.
Then I deleted that DB and selected 5 images (including the original one). The original one had 35 score now!
I guess it might be related to the fact that not only is level of detail important but also how tough it would be to distinguish that image from others.
That is my guess only tough.
Image scores are image specific and aren't affected by other images in the database.
You can currently get a score using the command line tool in the SDK (macOS example shown here):
augmented_image_cli_osx eval-img --input_image_path filename.png
@synthercat I guess the score difference you mentioned concerns particular image not the whole db, @fredsa should not the scores be the same then?
I believe what's happening is that the Augmented Image Database inspector is showing the quality score that was calculated when the image was added to the database (typically at database creation time), and is failing to update these values when the underlying imgtool is updated to calculate quality scores differently.
Workaround:
Most helpful comment
I guess it might be related to the fact that not only is level of detail important but also how tough it would be to distinguish that image from others.
That is my guess only tough.