Understanding the detector negative and positive overlap range [0 0.3]
1 visualización (últimos 30 días)
Mostrar comentarios más antiguos
Hi All,
I am trying to get the understanding of the [o o.3] range!
Based on my little knowledge, this range specifies that the bounding boxes over 0 0.3 range is considered outliers and systematically discarded and everthing below determines the objects location and score?!!?!?!?!?
Can someone assist me with a simple breakdown for this please?
I read in the matlab documentation below but still this has me a bit confused????!!!
does this range toggle the score output?
I also tried not suppressing the output to see what actually happens to get a holistic view on the operations
% 'PositiveOverlapRange' A two-element vector that specifies a range of
% bounding box overlap ratios between 0 and 1.
% Region proposals that overlap with ground truth
% bounding boxes within the specified range are used
% as positive training samples.
%
% Default: [0.5 1]
%
% 'NegativeOverlapRange' A two-element vector that specifies a range of
% bounding box overlap ratios between 0 and 1.
% Region proposals that overlap with ground truth
% bounding boxes within the specified range are used
% as negative training samples.
%
% Default: [0.1 0.5]
Thank you in advance for acknowledging my question
2 comentarios
Image Analyst
el 11 de Mayo de 2020
I have no idea where those comments came from, other than it's some function you have.
How are you getting scores and bounding boxes? Are you calling regionprops() or something? Is this an Image Processing thing, a Computer Vision thing, or a Deep Learning thing? I have no idea. Why would bounding boxes overlap, and what does it matter if they do? Can you show an image with the bounding boxes overlaid on it?
Respuestas (1)
Harsha Priya Daggubati
el 14 de Mayo de 2020
Hi,
This overlap range might make more sense if you are aware about how object detectors work. This particular trainECNNObjectDetector method is used to train an object detector to detect multiple objects in a given image. Positive Overlap Range suggests to treat the region under consideration (a sub-part in a training image) as positive (i.e. containing an object of importance) by calculating Intersection over Union (IoU) with the ground-truth is in the supplied range. Similarly treats as a negative example, when the IoU is in NegativeOverlapRange given.
Hope this helps you!
0 comentarios
Ver también
Categorías
Más información sobre Image Processing Toolbox en Help Center y File Exchange.
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!