1
$\begingroup$

I have stereo pairs (left, right) images of concrete cracks. I want to measure the length of the crack from those image pairs. Which neural network is appropriate for measuring object dimensions from stereo images?

Note: I am insisted to use the NN-based technique only.

$\endgroup$
  • $\begingroup$ Hi and welcome to this site! What do you mean by "images of a site"? Which site? Maybe you can provide an example of a pair of images with the object whose size you want to measure. I think that would clarify your problem. $\endgroup$ – nbro Nov 19 at 2:59
  • $\begingroup$ Thanks for the intimation. I have edited my description. $\endgroup$ – ovezn Nov 19 at 8:25
3
$\begingroup$

Is the image taken from a constant distance?

If yes, you'd need to scale the images to the same dimensions first of all. For few images say 100-500 images (more the better) you'd need to label the dataset by proper scaling.

Once labeled, use it to train a CNN (Although best would be training a ResNet). Once trained with decent accuracy, test it for the rest of your dataset.

I did something similar for one of my projects, check it out if you want to here.

$\endgroup$
  • $\begingroup$ Thank you for your information and reference. Will check it. $\endgroup$ – ovezn Nov 20 at 3:08
2
$\begingroup$

If you have stero pairs, and you can identify the objects in the scene, you do not need a neural network, you can just use triangulation.

If you need to identify which objects in the scene are the same, you have an image segmentation problem. Depending on your problem and the amount of data you have access to, you may be able to use simple techniques like clustering-based segmentation, or you may be able to use NN-based techniques, like Mask R-CNN.

$\endgroup$
  • 1
    $\begingroup$ I am supposed to work with NN only. My application is measuring the crack length from the underwater concrete images. I am aware of doing semantic segmentation using Mask R-CNN. But I don't know how to use it to measure the length. Could you please provide any suggestions? $\endgroup$ – ovezn Nov 19 at 8:18
  • 1
    $\begingroup$ @ovezn If you have stero images of the same object, taken with a stero camera, you can use triangulation to determine depth (i.e. the distance of the object from the camera). As soon as you know that, you can determine the length of an object in the picture. So the only thing you really need to do here is identify the object in the picture that is a "crack". That last part is an image segmentation task. Although you could try to solve the rest of the task with a neural network too, that's likely to be both harder, and less effective, than doing it properly by exploiting the depth. $\endgroup$ – John Doucette Nov 19 at 15:26
  • 1
    $\begingroup$ Many thanks for your information. $\endgroup$ – ovezn Nov 20 at 3:10

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.