Beamr's video compression tech improves AI model accuracy for autonomous vehicle safety by 30.7%.

Beamr Imaging Ltd. demonstrated that training AI models on video compressed with its patented Content-Adaptive Bitrate (CABR) technology reduces depth estimation errors by 30.7% for vulnerable road users like pedestrians and motorcyclists. This approach also cuts video file sizes by 35.2%, lowering storage and network costs while enhancing AI model resilience. The findings suggest that adaptive compression can be a valuable tool in AI development pipelines, especially for autonomous vehicles and machine vision applications, rather than just a cost-saving measure. Beamr plans to continue validating this technology across various AI workflows.