Live: How to vectorize line drawings with machine learning

MARCH 6, 2022

Back in September 2021—right at the start of Season 2 of the live stream1—I managed to get the so-called Virtual Sketching framework to work. We edited and released a twenty-minute video from that stream with an overview of the paper, its drawing representation, and a hands-on tutorial on how to build a Python notebook in Google Colab to get the algorithm to work.

Among other tasks, the machine learning program lets you vectorize line drawings as individual strokes—polylines, I believe—and save the vectorized sketches as SVG or export them as GIF animations.

Over the past streams,2 I continue playing with the program and managed to produce animations of my own drawings, as you can see here.

We'll continue to learn more about how SVGs and GIF animations are created from the model's NPZ3 prediction output.

I hope you'll Join us live on Thursday, March 10, at 11:30 ET! =)

  1. That was Live 39

  2. In Live 62 and Live 63

  3. "The .npz file format is a zipped archive of files named after the variables they contain. The archive is not compressed and each file in the archive contains one variable in .npy format." numpy.savez. NumPy.