Data Science

AI Can Now Create Websites From Drawings

Mar 22, 2018
By Nefi Alarcon
Discuss (0)

AI-Generated Summary
  • A new deep learning model converts hand-drawn wireframes into functional HTML websites, compressing a workflow that normally takes weeks.
  • Developer Ashwin Kumar trained the neural network on 1,750 screenshots of synthetically generated websites and their source code using NVIDIA Tesla V100 GPUs on Amazon Cloud with the cuDNN-accelerated TensorFlow framework.
  • The approach targets the bottleneck in traditional design workflows where product managers, designers, and engineers iterate through research, mockups, and code implementation.
  • Kumar notes the model currently has limitations but can improve by training with additional elements such as images, drop-down menus, and online forms.

Next Steps

Powered by NVIDIA Nemotron. AI-generated content may summarize information incompletely. Verify important information. Learn more

A new deep learning model allows users to create working HTML websites from hand-drawn wireframes, creating a solution for a process that normally takes weeks and involves multiple stakeholders.
The developer, Ashwin Kumar, recently participated in Insight’s Data Science Fellowship Program, which aims to bridge the gap between academia and data science. There, he developed a unique and breakthrough technology that allows anyone to build a website with just a simple sketch.
Using NVIDIA Tesla V100 GPUs on the Amazon Cloud and the cuDNN-accelerated TensorFlow deep learning framework, Kumar trained his neural network on 1,750 screenshots of synthetically generated websites and their relevant source code.
“My goal at Insight was to use modern deep learning algorithms to significantly streamline the design workflow and empower any business to quickly create and test webpages,” Kumar wrote in a Medium post.

A typical design workflow: A product manager performs research and creates a list of the specification needed. Designers than take those lists and create mockups. Engineers implement those designs into code.

Kumar mentions that the current design process is slow and can quickly turn into a bottleneck,  preventing startups and small businesses from getting off the ground.
He concedes that his model has some limitations, but explains that by training his neural network with additional elements such as images, drop-down menus, and online forms, it will improve the algorithm.
Kumar now works as a deep learning scientist at Mythic, a startup that aims to bring AI to every connected device.
Read more >

Discuss (0)

Tags

Data Science | cuDNN | News | Machine Learning & Artificial Intelligence

About the Authors

Nefi Alarcon
About Nefi Alarcon
Nefi Alarcon is a senior executive communications manager on NVIDIA's leadership team. He has years of media relations and communication experience, and has previously worked at Google, Mozilla, and CNN. He received his bachelor's degree in Journalism from George Washington University.

Comments