Recognizing images without neural networks
category: code [glöplog]
So you've reinvented OCR, but by a factor of thousands slower. And still you're only able to recognize one letter with fixed width, without even determining how many lines you need to represent it. There are already lots of computer vision research papers dealing with text recognition that outperform this approach.
In what way is this better? Oh and don't come with the "someday we will recognize every image with this" bullshit. You've been bragging about an AI revolution for over 5 years and still got nothing notable (meaning lots of people except yourself use and recommend it) to show. Alles nur heiße Luft.
Other than that, nice little experiment. Looks like fun, but that's all. Even so, assuming you're not just toying around, what's the problem you're trying to solve? Sure, an image recognition from an array of pixels to a set of data. But what does that data mean? Do you extract useful information from your list of 2d points? I fail to see the purpose of all this.
In what way is this better? Oh and don't come with the "someday we will recognize every image with this" bullshit. You've been bragging about an AI revolution for over 5 years and still got nothing notable (meaning lots of people except yourself use and recommend it) to show. Alles nur heiße Luft.
Other than that, nice little experiment. Looks like fun, but that's all. Even so, assuming you're not just toying around, what's the problem you're trying to solve? Sure, an image recognition from an array of pixels to a set of data. But what does that data mean? Do you extract useful information from your list of 2d points? I fail to see the purpose of all this.
