Does this support COCO datasets for instance segmentation?

The first item in the fashionpedia dataset has the following data:

image_id (int64)	image (image)	width (int64)	height (int64)	objects (sequence)
23		682	1,024	{ "bbox_id": [ 150311, 150312, 150313, 150314 ], "category": [ 23, 23, 33, 10 ], "bbox": [ [ 445, 910, 505, 983 ], [ 239, 940, 284, 994 ], [ 298, 282, 386, 352 ], [ 210, 282, 448, 665 ] ], "area": [ 1422, 843, 373, 56375 ] }

What is the format of the bbox? It appears to be x1,x2,y1,y2 format
If that is indeed the format of the bbox, how is the area being calculated here?

My calculations don’t match the examples' at all.
For the first bounding box:

• Width = 505.0 - 445.0 = 60.0
• Height = 983.0 - 910.0 = 73.0
• Area = Width * Height = 60.0 * 73.0 = 4380.0

For the second bounding box:

• Width = 284.0 - 239.0 = 45.0
• Height = 994.0 - 940.0 = 54.0
• Area = Width * Height = 45.0 * 54.0 = 2430.0

For the third bounding box:

• Width = 386.0 - 298.0 = 88.0
• Height = 352.0 - 282.0 = 70.0
• Area = Width * Height = 88.0 * 70.0 = 6160.0

For the fourth bounding box:

• Width = 448.0 - 210.0 = 238.0
• Height = 665.0 - 282.0 = 383.0
• Area = Width * Height = 238.0 * 383.0 = 91154.0

Those don't match [ 1422, 843, 373, 56375 ] respectively.