Annotated Data is black

[arnav2]

I don't get the annotated images. They just appear black ?

[ianzur]

Same, I would love to know what an example annotated image should look like for training on a dataset of my own.

[jay-thakur]

+1 . is there anyone who have annotated own image ? would love to know what tool do you use ? & how does it look like ?

[svanbodegraven]

Yep. Why are they black @alexgkendall

[carbocation]

They aren't meant to be interpreted perceptually. Each different label is assigned to a slightly different shade of black in ascending order. For example, some of the image is #010101 while other parts are #040404. These will both look black to your eyes, but you can re-color however you like and then you'll see that these different blacks are the annotations you're looking for.

[EtagiBI]

In my view, it's a VERY unclear way of annotating images.

[SangamSwadiK]

@carbocation Thanks for the explanation, basically, the person who did that, used the same way of annotation, but instead of using different colors for different objects , he has used different shades of black.

[dancassin]

I believe the reason why you're seeing black has more to do with how GitHub renders the images than the person who created them. These were likely all created with Python (or similar language) assigning a 0,1,2 etc (a different number for each appropriate class) to each pixel. If you import the images into Python and assign a cmap='gray', you'll see them more like a photo negative.