Add BackgroundColorAnnotator

[capjamesg]

Description

This PR adds a new sv.BackgroundColorAnnotator that highlights all regions outside of a list of bounding boxes or masks.

Type of change

• New feature (non-breaking change which adds functionality)

How has this change been tested, please provide a testcase or example of how you tested the change?

You can test this feature using the following Colab:

https://colab.research.google.com/drive/18GLw4TR-au9DUmAeWU30M7BFJ_00BlxO?usp=sharing

Any specific deployment considerations

N/A

Docs

N/A

[LinasKo]

Let's explain in more detail what happens. For folks relying on VSCode docs, this it the #1 most important place for documentation.

I'd like to know:

• what this does. (draws outside boxes and masks)
• how it treats mask and xyxy (boxes).

[LinasKo]

I really like this one - it'd be a key component in a project idea I have.

@capjamesg, I've made some comments. Could you also add a default image to the Colab?
Something like !wget https://media.roboflow.com/notebooks/examples/dog.jpeg should do the trick.

I'll do a deeper code review later on, but it generally looks okay.

[SkalskiP]

for x1, y1, x2, y2 in detections.xyxy.astype(int) should work as well and be shorter.

[SkalskiP]

Hi, @capjamesg 👋🏻 would you have time to respond to the comments above?

[capjamesg]

@LinasKo @SkalskiP I have responded to your feedback and added documentation to the detections Annotators documentation page.

[LinasKo]

Hey @capjamesg,

Could we get an expanded Colab, please?

Here's the tests I'd like to do:

1. Test on image with many detections (you have this already)
2. Test with no detections (you can call sv.Detections.empty()
3. In-place call of the annotator:

# Instead of:
annotated_image = bounding_box_annotator.annotate(
    scene=image, detections=detections)
    
# Call:
bounding_box_annotator.annotate(
    scene=image, detections=detections)
    
# This should still annotate the image!

Also, since I've seen different behaviours recently, it would be nice to check if both in-place and not in-place work on PIL images. You can use sv.cv2_to_pillow(image) before passing to the model. (or just load with PIL)

[capjamesg]

@LinasKo See my revisions to the Colab and code.

[LinasKo]

Close, but to match the behaviour of annotators, it should work in-place for numpy images. I mean, it should modify the image you pass in: https://colab.research.google.com/drive/1SP4QyD_MgvWz9UI6qdkjAIaF74eRoUTy?usp=sharing

Can you modify and return the same image the annotator gets as an argument, instead of returning a copy with Image.fromarray(colored_mask[:, :, ::-1])? np.copyto may or may not be useful - I'm not sure.

Let me know if it gets too much - I'm a bit busy, but could help out 😉

[SkalskiP]

@LinasKo / @capjamesg, what do you think about renaming it to BackgroundFillAnnotator or BackgroundOverlayAnnotator?

[LinasKo]

Yesss, I like the new name much more

[SkalskiP]

@LinasKo which one? :)

[LinasKo]

I misread; I thought you mentioned the previous one too.

BackgroundOverlayAnnotatormakes the most sense to me.

[SkalskiP]

@LinasKo, is there any chance you could propagate that name change in this PR? We wouldn't need to ask @capjamesg to work on it once again.

[LinasKo]

On it.

[LinasKo]

Okay, I can take this one over.

The return np.copyto is incorrect, but we're a millimeter away from the solution 😆

@capjamesg, thanks a lot! It's an awesome annotator to have in our toolbox.

[LinasKo]

Changed the name, return type.

Tested here: https://colab.research.google.com/drive/1SP4QyD_MgvWz9UI6qdkjAIaF74eRoUTy?usp=sharing