0
$\begingroup$

  1. I am thinking of choosing a computer vision project for my school project(detect crack on surface) and the duration I have is roughly 4 months. With no prior knowledge in neural network, is matlab computer vision application consider to be "user friendly" to beginner?

  2. When detecting a defect by comparing standard image to product image that work by checking the RGB contents of pixels to pixels. Does such method still involve the use of neural network(ie. the RGB information of a pixel can be set not to exceed certain value, otherwise it will mark that pixel as defect)?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

2 Yes, cracks and in general defects detection is a task performed with neural networks and deep learning, but depending on your specific use case it might be an overkill. You talk about pixel to pixel comparison, if your data are images taken from a production line with standardize equipment that capture images of the same piece always with same lighting, angle, etc then you'll probably want to investigate analytic methods for more explainable and efficient results. Neural nets become almost compulsory when the images to compare differs a lot between each other, making hard for an expert to came up with analytic rules that are always applicable.

1 I never played a lot with matlab and its deep learning libraries, but I know that at the very least it allows to use models trained with python libraries like tensorflow, this might be an advantage since at least you know you can search models and repositories not exclusively written in matlab.

Improve this answer
$\endgroup$

Not the answer you're looking for? Browse other questions tagged .