Binary image classifier for logos

Question

I am currently looking for a possibly fast way to create (or use an already existing) classifier for images. The main goal is to classify the images whether they are (mainly) a logo or not. This means that I am not interested in recognizing the brand/name of the company. But instead the model would have to tell how possible it is that the image is a logo.

Does such a categorizer already exist? And if not, is there any possible solution to avoid neural networks for this task?

Thanks in advance.

Answer 1

I am not sure about the existence of this project, but I have a couple of ideas that can work for this without neural networks. I think as a convention neural networks would be much easier but I think it might be done K-means algorithm or by a clustering algorithm. I have imagined like if logo data are in the same area and image data are in another same area, they can be clustered.However, I haven't done it sth like that before but theoretically, it seems logical

Answer 2

As you have seen there isn't ay label in our algorithm.