Enthusiasm over the potential social impact of machine learning is not just limited to big gigs; it has the potential to solve many of the regular human issues. By using machine learning tools, enormous gains can be made by common people. But some of the most important issues generally faced today fall within the cracks of not knowing the right way to use any of these tools.
For addressing this issue, Christopher Hesse has created a web based tool named as edge2cats. For translating one image to another, this machine learning tool applies Google’s Tensorflow. It uses training data provided by a database of more than 2,000 stock images of cats. The limitations set upon this tool forces it to create images relating to cats. The stock images are provided to identify edges and fill in simple line art so that it can approximately create an image of a realistic cat.
Hesse indicates that the results of some of the pictures come out pretty horrifying. That is because it’s pretty clear what animal faces actually look like and anything featuring eyes in particular looks especially creepy. This tool can take your basic line art come out in a unique style among digital artists. Previously similar project by Hesse included variants that could color in handbags, shoes, and build facades with not much scary outcomes.
Previously it was impossible to imagine such things but AI has provided a framework for people to interact effectively with machine learning models. Machine learning techniques and human skills complement each other and this effort is a good example that people can interact with machine learning systems without having an in-depth knowledge about its techniques.
Via: Tech Crunch