Answer the question
In order to leave comments, you need to log in
How to implement a network for generating a geometric image from a sketch?
To make it work like this, we
divide it into tasks:
Answer the question
In order to leave comments, you need to log in
Googled GAN, and although there are technically two networks, you can perceive it as one neuron, because they train together. And in general, there are no preliminary divisions into figures, colors, etc.
Neural networks will not help you if you do not know how to extract the original information from the image. And besides, where can you find several tens/hundreds of thousands of images for network training? Neuronki - this is not magic, to vzhuh and it's on its own. -_-
I would try to find all the pixel colors that occur and cluster them, but the matter is complicated by the fact that the number of colors used is not known in advance. After that, it will be possible to select each blackness separately by comparing the color of the pixel with the center of a particular cluster, and then separating individual fragments through the search for connected components .
Next, you can take this algorithm as a basis .
Didn't find what you were looking for?
Ask your questionAsk a Question
731 491 924 answers to any question