sjv

6 ms·

1.
by sjv 6y ago
I implemented the network from scratch in vanilla javascript. The file is just a custom binary file containing the names, shapes, and weights of the network parameters. If you have more specific questions, feel free to ask them.
2.
by sjv 6y ago
The dataset is from the IAM On-Line Handwriting Database. I've linked my github, which has my email, on my profile.
3.
by sjv 6y ago
Outputs are sampled from a probability distribution, and increasing the legibility effectively concentrates probability density around more likely outcomes. So you're correct that it's just altering variation. The general techni
4.
by sjv 6y ago
Hey HN — This is a simple web app that allows you to enter text and watch as an in-browser neural network generates the corresponding handwriting in real time. I’m happy to hear feedback or answer questions about how it works!
5.
Show HN: Generate realistic handwriting with neural networks [demo](calligrapher.ai)
66 points by sjv 6y ago|19 comments
6.
by sjv 6y ago
The value is almost entirely as a technical demo of current machine learning techniques. Handwritten fonts will always be a more reliable way to create synthetic handwriting, so I don't see this as something that will be used as anyth
7.
by sjv 6y ago
Hey HN — This is a simple web app that allows you to enter text and watch as an in-browser neural network generates the corresponding handwriting in real time. I’m happy to hear feedback or answer questions about how it works!
8.
Show HN: A realistic handwriting generator, directly in the browser [demo](calligrapher.ai)
4 points by sjv 6y ago|4 comments
9.
by sjv 6y ago
Yes, it's mostly based on that. Primary difference is that this model adds an inference model (VAE style), which allows you to a) sample style vectors from a latent distribution b) do so more efficiently than the priming mechanism des
10.
by sjv 6y ago
It's not intentionally slowed down. The sampling process is computationally heavy and you'll notice improved performance when running it on better hardware. There are many algorithmic ways to speed up neural network computation
11.
by sjv 6y ago
The model was trained on a relatively small dataset which has few occurrences of the letter x, almost none of which are in cursive. As you've found, asking the model to write out an x in a cursive style can totally derail the generati
12.
by sjv 6y ago
Non-statistical methods of imitating handwriting are always fairly easy to identify as synthetic due to the uniformity of strokes and unnatural kerning (among other things). Even methods which incorporate random variations don't hide t
13.
by sjv 6y ago
This problem was more or less solved with neural networks in 2013, see https://arxiv.org/abs/1308.0850 . I've implemented a similar model in javascript that you can interact with: https://seanvasquez.co
14.
by sjv 6y ago
Perhaps something like this? https://seanvasquez.com/handwriting-generation
15.
Show HN: Handwriting generation with an in-browser recurrent neural network(seanvasquez.com)
2 points by sjv 7y ago|0 comments