1
$\begingroup$

I have a dataset which contains information about objects from 20 different categories. These objects are manually labeled on 100 images. The objects are located in 3 rows on a square image, there can be multiple instances of the same class object. For a given object, I have the following data:

  • its class
  • bounding box center coordinates
  • bounding box size
  • information whether the object is in the correct position or not

The ultimate goal is to have a classifier, which determines if the position of the detected object is correct. I am not sure if I should treat a label or "correctness" as an output, what would you suggest? Also, what are the best candidates for classifiers in case of this data, given the following:

  • Each training image (with labelled objects) is warped by a small angle, perspective, translations etc.
  • There is one correct arrangement of objects on the image

Please look below for simplified illustration of the data: number - label, bold - incorrect position

===================

1 1 1 3 3 5

6 7 1 1 4 4

9 5 5 5 9 7 7 7

===================

$\endgroup$
1
$\begingroup$

It might actually be called Intersection Over Union.

So how do you tell if your object detection algorithm is working well? In this video, you'll learn about a function called, "Intersection Over Union". And as we use both for evaluating your object detection algorithm, as well as in the next video, using it to add another component to your object detection algorithm, to make it work even better. Let's get started. In the object detection task, you expected to localize the object as well. So if that's the ground-truth bounding box, and if your algorithm outputs this ...

| improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.