Labels provided in JSON format not recognized

We wanted to seek suggestions and/or pointers regarding a problem we encountered with JSON file based labels provided within a data set when loading to Roboflow. First I wanted to mention that I am beginner with the Roboflow although have extensive experience in coding and HPC. I am just trying to help with an effort to identify damaged buildings in the recent disaster in Turkey due to back to back two strong earthquakes in the same region. We started getting the recent satellite images of the region. So, we wanted to build our workflow on Roboflow. We first wanted to train a model using available datasets.
Hence, we are trying to use a dataset from an earlier study ([1911.09296] xBD: A Dataset for Assessing Building Damage from Satellite Imagery) which has labeled earthquake dataset with before and after the earthquake images to train models using Roboflow environment (thanks to kind and generous support from them).
This dataset sets comes with PNG images for pre- and post-event. Also, there is a labels directory which contains JSON files that I presume are for the building polygons provided as labels. Same JSON file exists for post-event images also.
Here is a short excerpt from one of them:
$ head …/…/xBD/train/labels/mexico-earthquake_00000001_pre_disaster.json | more
{“features”: {“lng_lat”: [{“properties”: {“feature_type”: “building”, “uid”: “f61736db-c71a-4eac-95ab-704909c1a133”}, “wkt”: “POLYGON ((-99.22391262995048 19.3327103251249, -99.22399703307177 19.33267259856737, -99.2240258747197 19.33268538911176, -99.22403039481853 19.33269735447365, -99.22408170520377 19.3326733495849, -99.22413409480619 19.33278016836537, -99.22391284437903 19.33286765690071, -99.22391262995048 19.3327103251249))”}, {“properties”: {“feature_type”: “building”, “uid”: “528c7a1e-46ca-4795-bcdd-92f1ec30889a”}, “wkt”: “POLYGON ((-99.22392120853468 19.33264649076796, -99.22394978080557 19.33262876972653, -99.22394153057537 19.33261561822526, -99.22391512077405 19.3326304575019, -99.22392250668834 19.33264525527029, -99.22392120853468 19.33264649076796))”}, {“properties”: {“feature_type”: “building”, “uid”: “fed2b4e0-d48e-42e3-a555-5d5f1faf3585”}, “wkt”: …

% ls -als …/…/xBD/train/images/mexico-earthquake_00000001_pre_disaster.png
3888 -rw-------@ 1 egel2 UNL-AD\Domain Users 1988853 Feb 12 13:39 …/…/xBD/train/images/mexico-earthquake_00000001_pre_disaster.png

I naively tried loading both matching PNG image and JSON files together in the same upload session with the expectation that Roboflow might detect the JSON based file as labels. However, it didn’t or there is something I am missing, which is the reason I wanted to ask for suggestions.
Is there a quick way to have the label information from these JSON files imported to Roboflow or do I need to convert to another format? Any suggestions will be very helpful.

Thank you in advance for your suggestions and help.

Aytekin Gel, Ph.D.

These JSON files don’t look like they’re in a supported annotation format.

They appear to be latitude/longitude coordinates, not image coordinates so it seems unlikely we’d be able to add support for importing them since they don’t have enough information to convert them (but I don’t have the full dataset available to dive in deeper; only going on what you put in this post).