The EPO published patent application EP3942467A1 filed by Google LLC on March 25, 2026. The invention relates to systems and methods for machine learning-based image compression that dynamically adjusts compression settings based on user preferences, utilizing neural network architectures including GCNNs to predict optimal quality-file-size tradeoffs. The application covers multiple IPC classifications including G06N 3/0464, G06F 16/583, G06T 9/00, and G06V classifications covering image processing, recognition, and pattern analysis.

Technology companies developing image compression, delivery systems, or media processing applications should review this patent publication to assess potential exposure. The broad IPC classification coverage suggests applicability to ML-based image and video processing systems. Upon patent grant, companies practicing similar ML-based compression techniques may need to evaluate licensing requirements or design-around options.