Question about dataset file split.json

[sshiningg]

Thank you for your excellent work! I have a small question about the released dataset.
Why is there always one fewer file under the 'extra' tag in the split.json file compared to the number of images that should be categorized as 'extra' in the images file?

For example in the spot scene, there are 10 images under the 'extra' tag in the split.json:

but there are 11 images in the images file that should be categorized as ‘extra’:

This issue also occurs in other quantitative assessment scenarios.

[rwn17]

Thank you for bringing this to our attention. You are correct. After reviewing the data, I can confirm this inconsistency occurred during the very early generation of the 'extra' evaluation index.

To ensure fair comparison, all methods (including on-the-go) use the same data split. Please use the provided split file for evaluation and report the metrics based on that.

Best regards,
Weining