Utility for merging detection results from sliced coco dataset

[rohansaw]

I used the coco slicing utility to slice my coco-format dataset into 512x512px patches with an overlap of 0.25. I then used a custom object detector to perform inference on these patches and have the results as a coco-style results json for the patches.

Is there a utility to combine the results, so that the result patches of an image are merged into detection results on a single and can be evaluated against the original coco annotations? I didn't find a direct utility for this in the repository.

Thank you!

[Trotts]

Unsure if you were able to sort this, but I had a similar request. See #1069 (comment) for my code.

(sorry I am many months after you asked this -- perhaps it will be useful for people who stumble upon this question)