DEEP REFACTOR / TechnologyLet’s think further

THE VISUAL PIPELINE

Reconsider
every layer.

AI video processing is a connected problem: representation, temporal context, compression and deployment. Our research direction follows those connections.

INTERACTIVE REPRESENTATION STUDYDR / LAB
+
+
100%

An illustrative downsampling study using generative motion. This is not a compression benchmark or a demonstration of a Deep Refactor codec.

01

Find the useful structure.

Spatial patterns and relationships between frames provide opportunities to represent visual information more efficiently. The first question is what the application needs to preserve.

Read the research