The computer calligraphy is same as human
By , www.dailymail.co.uk
Computer calligraphy. For some, the personal touch that comes from a handwritten letter is being lost due to the use of email, text messages and computers to send messages. But computers themselves may soon have their own personalised handwriting with the help of networks that mimic the structure of the human brain.
Computer scientists have created a system known as a ‘recurrent neural network’ that can generate its own handwriting. The program can learn how to create handwriting-like text by examining the movement of a pen as real people write. The system can then use this to mimic the handwriting styles and even create some of it is own.
While the system still needs to have text inputted – it cannot yet produce its own content – it can produce authentic looking handwritten notes.
Some of the words run into each other, others are illegible and the lines themselves can slant alarmingly downhill. At one point the system also appeared to suffer some something many students will likely identify with – the virtual pen seemed to drift off the page as if it had fallen asleep mid-flow. It will even scribble out mispelt words and errors, much like a real human.
The system was developed by computer scientist Alex Graves, a computer scientist at the University of Toronto to demonstrate how a neural network can be trained. He said:
‘The resulting sequences are sufficiently convincing that they often cannot be distinguished from real handwriting. Furthermore, this realism is achieved without sacrificing the diversity in writing style.’
The computer can generate one of five styles that mimic the writing style of a human or it can create an entirely random style of its own.
computer calligraphy software
An online prototype of the system allows users to input 100 characters at a time, meaning longer pieces of text can often look slightly variable.
However, it is possible to make the handwriting more legible and neater using a ‘bias’ slider that improves the handwriting.
Recurrent neural networks are being developed as part of artificial intelligence research by mimicking the connections of neurons by connecting each ‘unit’ in it to every other unit. It connects the basic units in a cycle which allows it to deal with ‘fuzzy’ information like handwriting.
Mr Graves, who has published a paper on his work on arXiv.org, said he hoped to adapt the handwriting system to more realistic voice generation too. He said:
‘(It) which is likely to be more challenging than handwriting synthesis due to the greater dimensionality of the data points. It would also be interesting to develop a mechanism to automatically extract high-level annotations from sequence data. In the case of handwriting, this could allow for more nuanced annotations than just text, for example stylistic features, different forms of the same letter, information about stroke order and so on.’