loss of annotations

[withTai]

Thanks for you brilliant work!
But I found several issues with the annotations in this dataset.

1. Severe annotation coverage gapMost images are missing corresponding JSON annotation filesExample from Blade_1/1_1: 17 images but only four have JSON annotations (~23.5% coverage)
   Example from Blade_15/1_1: 52 images but only 8 have JSON annotations (~15.4% coverage)

2. Label format inconsistency
   Case sensitivity issues: "surface;contamination;other" vs "Surface;contamination;dirt"
   Inconsistent classification hierarchy: some labels use 2 levels (e.g., "add-on;serration;OK"), others use 3 levels (e.g., "leading edge;erosion;coating or LEP only")
   And most importantly, it seems that the bboxes are not in one-to-one correspondence.

Could you please:

Provide the complete annotations for all images?
Standardize the label format across all annotations?

Thank you for your work on this dataset.