CODEWITHSUNDEEP
pythonimage-processing
Ryan
Ryan

Reputation: 10109

Method to modify bounding box co-ordinates

I have 8 points for every image in my dataset, (x1, y1, x2, y2, x3, y3, x4, y4).

My objective is to train Retina net, but the repository requires that the format be like

path/to/image.jpg,x1,y1,x2,y2,class_name

they only demand 4 points, something like this /data/imgs/img_001.jpg,837,346,981,456,cow

My 8 points for one image look like this - [[220, 129], [1028,113], [1098,684], [206,706]

Is there any way i could convert my 8 points to 4.

Thanks in advance

Upvotes: 0

Views: 84

Answers (1)

AdamKG
AdamKG

Reputation: 14081

Assuming the four pairs are just four x,y points in the image that should be in the bounding box,

>>> vals = [[220, 129], [1028,113], [1098,684], [206,706]]
>>> bbox = [min([x[0] for x in vals]), min([x[1] for x in vals]),
...         max([x[0] for x in vals]), max([x[1] for x in vals])]
>>> print(bbox)
[206, 113, 1098, 706]
>>> imgpath = 'path/to/image.jpg'
>>> print('{},{},{}'.format(imgpath, ','.join(map(str, bbox)), 'clsname'))
path/to/image.jpg,206,113,1098,706,clsname

Upvotes: 1

Related Questions