Json format for use with 3D Pose motion bert

[CrisTTT]

Hi, I'm trying to inference motion bert in the wild. I am having issues with the JSON format that I should use for this.

Below is the json format i am currently using. I can't find an example of a good JSON used for this in order to reformat it to the correct format.

{
"annotations": [
{
"frame_id": 0,
"keypoints": [
256.0015869140625,
455.171630859375,
0.0,
258.3621826171875,
452.07177734375,
0.0,
264.28759765625,
448.86700439453125,
0.0,
269.7962646484375,
457.30731201171875,
0.0,
301.1651611328125,
450.5689697265625,
0.0,
272.2452392578125,
496.71942138671875,
0.0,
332.996337890625,
473.644287109375,
0.0,
259.823974609375,
527.5805053710938,
0.0,
360.90948486328125,
497.08831787109375,
0.0,
254.77301025390625,
496.04486083984375,
0.0,
375.521484375,
638.1166381835938,
0.0,
320.78118896484375,
609.2880249023438,
0.0,
365.06146240234375,
593.7268676757812,
0.0,
323.9300537109375,
698.5770874023438,
0.0,
357.780517578125,
632.9181518554688,
0.0,
329.4959716796875,
797.6443481445312,
0.0,
388.8597412109375,
688.8500366210938,
0.0
],
"bbox": [
0,
0,
478,
850
]
},
{
"frame_id": 1,
"keypoints": [
257.859619140625,
454.60906982421875,
0.0,
259.45166015625,
451.2744140625,
0.0,
266.6239013671875,
447.96405029296875,
0.0,
270.135498046875,
457.2606201171875,
0.0,
302.27679443359375,
450.3916015625,
0.0,
273.5531005859375,
496.06805419921875,
0.0,
333.98193359375,
473.67254638671875,
0.0,
259.32861328125,
526.0870971679688,
0.0,
361.90771484375,
499.54449462890625,
0.0,
254.3492431640625,
494.46270751953125,
0.0,
381.62225341796875,
648.5664672851562,
0.0,
320.60833740234375,
609.2427368164062,
0.0,
364.91314697265625,
594.4885864257812,
0.0,
324.10009765625,
698.9760131835938,
0.0,
356.1573486328125,
631.1774291992188,
0.0,
327.9652099609375,
795.2775268554688,
0.0,
389.0799560546875,
689.3659057617188,
0.0
],
"bbox": [
0,
0,
478,
850
]
},

[CrisTTT]

vis_climb-trim.json
Here is an example of a JSON I want to use. The error I am encountering for:
python infer_wild.py --vid_path climb-trim.mp4 --json_path vis_climb-trim.mp4.json --out_path output_dump

is the following:
Traceback (most recent call last):
File "infer_wild.py", line 61, in
wild_dataset = WildDetDataset(opts.json_path, clip_len=opts.clip_len, scale_range=[1,1], focus=opts.focus)
File "D:\motion\motionbert\lib\data\dataset_wild.py", line 95, in init
self.vid_all = read_input(json_path, vid_size, scale_range, focus)
File "D:\motion\motionbert\lib\data\dataset_wild.py", line 80, in read_input
kpts_all = halpe2h36m(kpts_all)
File "D:\motion\motionbert\lib\data\dataset_wild.py", line 51, in halpe2h36m
raise ValueError("Input keypoints do not match the expected format. Expected at least 20 keypoints.")
ValueError: Input keypoints do not match the expected format. Expected at least 20 keypoints.