my detector shows multiple bounding boxes .

hi everyone,
i have problem that my trained detector show multple bounding boxes on same object as shown in attached image .Inaddition to this iit shows bounding boxes only when threshold is set to 0.1.if i set 0.2 or more it does not show anyy box or label.

Answers (1)

I don't know what your detector does or how it operates or even what function you called. However maybe you can run through the bounding boxes and for those that have any overlapping area, take the bounding box of all the overlapped boxes.

9 Comments

My detector is yolov2 which detects pothole,speedbumps,pole and lane .i trained dataset on these classes.
So you are saying that I have to remove bounding boxes that overlap in training dataset.
I'm saying to construct the overall bounding box and get rid of the ones inside.
How can I construct on this test image detect by detector?
This is my code
img = imread("C:\Users\ZBook\Desktop\New folder\new road.v2i.darknet\train\2_jpg.rf.e03c022bb9ccc6987308e66d130f4db7.jpg");
[bboxes,scores,labels] = detect(detector,img,'Threshold',thrusholdvalue);
detectedImg = insertObjectAnnotation(img,"Rectangle",bboxes,labels);
imshow(detectedImg)
You forgot to attach "2_jpg.rf.e03c022bb9ccc6987308e66d130f4db7.jpg"
I change my image to check whether it detect or not
I place the attackrd image name in image to detect
img = imread("C:\Users\ZBook\Desktop\New folder\new road.v2i.darknet\train\2_jpg.rf.e03c022bb9ccc6987308e66d130f4db7.jpg");
Error using imread>get_full_filename
File "C:\Users\ZBook\Desktop\New folder\new road.v2i.darknet\train\2_jpg.rf.e03c022bb9ccc6987308e66d130f4db7.jpg" does not exist.

Error in imread (line 372)
fullname = get_full_filename(filename);
[bboxes,scores,labels] = detect(detector,img,'Threshold',thrusholdvalue);
detectedImg = insertObjectAnnotation(img,"Rectangle",bboxes,labels);
imshow(detectedImg)

Sign in to comment.

Asked:

on 19 Sep 2023

Commented:

on 20 Sep 2023

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!