Can this repository support change detection?

[yuchenwu73]

Search before asking

• [x] I have searched the X-AnyLabeling Docs and issues and found no similar questions.

Question

Can this code repository implement a change detection function? I.e. for two images, find out what makes them different and mark them using a mask. I think this is actually a semantic segmentation.

Additional

No response

[CVHub520]

Hi @yuchenwu73,

Thanks for opening this issue and for your interest in the repository.

Currently, the repository does not have built-in support for the change detection functionality you described (finding differences between two images and marking them with a mask).

You mentioned you think this might be semantic segmentation, but the task you described is specifically known as Change Detection. Just to clarify the distinction:

• Change Detection: Compares two or more images (often taken at different times) to identify areas where changes have occurred. The output is typically a mask highlighting these changed regions.
• Semantic Segmentation: Processes a single image to classify each pixel into a predefined category (e.g., car, road, building, sky). The output is a mask where each color/value represents a specific class.

So, your requirement clearly falls under change detection, which is different from semantic segmentation.

While direct support isn't available, the framework might be extensible depending on its design. For adding custom functionalities like this, you could potentially develop your own panel or tool. One approach could be to create a custom widget, perhaps by inheriting from QDialog (or a relevant base class from the project's UI framework), to implement the change detection logic and the user interface you need. This would allow you to integrate your specific change detection algorithm or workflow.

We hope this clarification helps. Let us know if you have further questions.