Could someone explain the structure of the semantic annotation?

Both the suadd_23_input_images-v0.1.zip and suadd_23_semantic_annotations-v0.2.zip contains a single folder with 1786 images.

Which float32 value of the image in semantic annotation represents what?

1 channel (grayscale) image with class label as value

Hi @vidit_agarwal

Here is the class map, from the baseline training repository.