Meta Research released a dataset for animating amateur drawings

Meta Research released a new dataset designed to help artificial intelligence models animate static drawings. The collection focuses on simple sketches often created by amateurs or children. It allows developers to train systems to recognize individual figures within a drawing and assign motion to them.

This release aims to solve the challenge of turning two-dimensional sketches into moving characters. The system identifies the limbs and joints of drawn figures to generate fluid movement. By providing this data, the company hopes to encourage further research into creative animation tools.

The current dataset is limited to specific types of amateur drawings. It does not handle complex illustrations or multi-layered artistic works. The accuracy of the motion generation depends heavily on the clarity and isolation of the figures provided in the sketches.