How to rearrange boxes automatically using nearest position in an image?

Question

I have two bound boxes. one for predict, which contains 3 rows the second for ground truth that contains two rows.

Ground truth boxes (x,y,w,h)

[[210, 98, 103, 88],
 [62, 75, 150, 120]]

Predicate boxes (x,y,w,h)

[[218, 104, 94, 69],
 [154 ,184, 34, 9],
 [67, 77, 143, 100]]

If I need to compute the bounding box overlap ratio using MATLAB function bboxOverlapRatio. The function evaluates the first-row box of ground truth with the first row in the predicate. Then second-row of ground truth evaluates with the second-row from predicate boxes. The function evaluates bound boxes using sequentially techniques row by row. The question is: What is a function that rearranges predicate boxes using the nearest position(place) or resorts boxes automatically by their appearance in the image? I want to rearrange to have a high ratio.

How can I do this if there is no MATLAB function?

Here is a sample code:

groundTruth = [210  98  103 88, 62  75  150 120]
predicate = [218 104 94 69, 154 184 34  9, 67 77 143 100];
ratio = bboxOverlapRatio(groundTruth,predicate)

I appreciate any help.

Answer 1

As per the documentation of the function bboxOverlapRatio,

Each (I, J) element in the output matrix corresponds to the overlap ratio between row I in bboxA and row J in bboxB.

For the function, your first argument (bboxA) is groundTruth and bboxB is predicate. Hence, in the ratio array, the element [0,1] represents the overlap ratio between grountTruth[0] and predicate[1]. Thus, for groundtruth[1], you can iterate through [1,0], [1,1], [1,2] and find the maximum value among them. Since [1,2] is the maximum, this indicates that for row 1 in groundTruth, row 2 in predicate has the highest value.

This might be useful - max, especially the section Return Linear Indices.