Beamr Cuts dSPACE RTMaps Video Logs by 31% While Preserving ML Accuracy
Beamr's Content-Adaptive Bitrate (CABR) compression cut dSPACE RTMaps video logs by 31% versus baseline encodes and by 97% versus uncompressed data while maintaining ML model accuracy within <2% mAP variance. The joint demonstration will be presented at the dSPACE user conference in Novi, Michigan, on April 21-22.
1. ML-Safe Compression Validation
Beamr and dSPACE conducted the first joint demonstration of CABR in the dSPACE RTMaps ecosystem, achieving a 31% file size reduction compared to baseline encodes and a 97% reduction versus uncompressed footage while preserving ML model accuracy within a <2% mAP variance for object detection tasks.
2. Impact on Autonomous Vehicle Pipelines
By applying ML-safe compression at the data logging stage, terabyte-scale multi-camera video data volumes are significantly reduced, lowering storage, transfer and infrastructure costs and accelerating AV development cycles without requiring changes to existing RTMaps workflows.
3. Future Testing and Deployment Plans
Beamr and dSPACE plan to expand their ML-safe compression trials to include video data simulation and hardware-in-the-loop testing, aiming to validate CABR across additional pipeline stages and ensure seamless integration into broader AV development environments.