Beamr (BMR) and dSPACE show ML-safe AV video compression with 31% size cut
Rhea-AI Filing Summary
Beamr Imaging Ltd. filed a Form 6-K highlighting a joint demonstration with dSPACE that validates “ML-safe” video compression for autonomous vehicle data inside the dSPACE RTMaps ecosystem. Testing on real-world sequences showed Beamr’s Content-Adaptive Bitrate (CABR) compression delivered 31% file size reduction versus baseline encodes and 97% reduction versus uncompressed data while preserving machine learning model accuracy.
The companies plan to extend ML-safe compression testing to additional stages such as video data simulation and hardware-in-the-loop testing. Beamr positions this capability as helping AV teams reduce data volumes and infrastructure demands without rebuilding existing RTMaps-based workflows.
Positive
- None.
Negative
- None.
Key Figures
File size reduction vs baseline: 31% reduction
File size reduction vs uncompressed: 97% reduction
Prior benchmark reduction: Up to 50% reduction
+2 more
5 metrics
File size reduction vs baseline
31% reduction
CABR compression on real-world AV video sequences in dSPACE RTMaps
File size reduction vs uncompressed
97% reduction
CABR compression vs uncompressed AV video data
Prior benchmark reduction
Up to 50% reduction
ML-safe video compression across AV pipeline in previous benchmarks
Object detection accuracy change
<2% difference in mAP
Impact of CABR on object detection mean Average Precision
Patents
53 patents
Intellectual property backing Beamr’s CABR technology
Key Terms
ML-safe compression, Content-Adaptive Bitrate compression (CABR), RTMaps, hardware-in-the-loop (HIL) testing, +2 more
6 terms
ML-safe compression technical
"validating, for the first time, compression for autonomous vehicle (AV) video data ... while preserving machine learning (ML) model accuracy"
ml-safe compression is a method of shrinking datasets or media so machine learning models can still read and learn from them without losing important signals or introducing bias. For investors, it matters because it can lower storage and bandwidth costs, speed model deployment, and reduce regulatory or operational risk by preserving model accuracy and fairness—think of compressing a photo but keeping the faces clear so automated recognition still works.
Content-Adaptive Bitrate compression (CABR) technical
"Beamr Content-Adaptive Bitrate compression (CABR) delivered 31% file size reduction compared to baseline encodes"
RTMaps technical
"Testing on real-world sequences processed through dSPACE RTMaps showed Beamr ... compression (CABR)"
hardware-in-the-loop (HIL) testing technical
"Beamr and dSPACE plan to extend ML-safe compression testing to additional stages, including ... hardware-in-the-loop (HIL) testing"
Hardware-in-the-loop (HIL) testing is a development and validation method where real physical components are connected to a simulated environment that mimics the rest of the system, allowing engineers to test how the hardware behaves under realistic conditions without building the full product. For investors it matters because HIL helps find design faults early, reduces costly field failures, speeds development and supports regulatory approval, which can lower project risk and preserve value — like test-driving an actual engine while the rest of the car is a simulator.
mean Average Precision technical
"For object detection tasks, CABR showed <2% difference in mean Average Precision"
Mean average precision is a single-number measure used to judge how well a ranking or recommendation system orders relevant items near the top; it averages precision scores at each point a correct item appears and then takes the mean across multiple queries. For investors, it matters when assessing tools that sort news, signals, or trade ideas — higher values mean the system more reliably surfaces the most important items first, like a well-organized playlist that puts your favorite songs at the top.
world foundation models technical
"Testing with world foundation models showed no measurable impact on AV captioning"
AI-generated analysis. How Rhea-AI works. Not financial advice.
FAQ
What did Beamr (BMR) announce with dSPACE in this Form 6-K?
Beamr announced a joint demonstration with dSPACE validating ML-safe video compression inside the dSPACE RTMaps ecosystem. Tests showed significant file size reductions for autonomous vehicle video data while preserving machine learning model accuracy, aimed at easing storage and processing burdens in AV development pipelines.
How much video file size reduction did Beamr’s CABR achieve in the dSPACE RTMaps tests?
Beamr’s CABR compression achieved 31% file size reduction compared to baseline encodes and 97% reduction versus uncompressed data. These results were obtained on real-world sequences processed through dSPACE RTMaps, while maintaining the accuracy of the machine learning models used on the compressed video data.
Why is ML-safe compression important for autonomous vehicle pipelines at Beamr (BMR)?
ML-safe compression lets AV teams shrink video data volumes without hurting model accuracy. Autonomous fleets generate terabytes of multi-camera footage, straining storage and slowing data transfer. Beamr’s approach targets these bottlenecks so teams can keep existing workflows while managing rapidly growing data requirements.
What future testing do Beamr and dSPACE plan beyond data logging?
Beamr and dSPACE plan to extend ML-safe compression testing to additional AV pipeline stages, including video data simulation and hardware-in-the-loop testing. This would broaden validation beyond data logging and replay, covering more of the development and verification workflow for autonomous vehicle systems.
How has Beamr’s CABR technology performed in previous benchmarks?
In prior benchmarks, Beamr’s CABR delivered up to 50% file size reduction for real-world and synthetic AV video data while remaining ML-safe. For object detection, it showed under 2% difference in mean Average Precision and no measurable impact on AV captioning when evaluated with two embedding models.
What markets and recognition does Beamr (BMR) highlight for its compression technology?
Beamr highlights use of its content-adaptive compression across media and entertainment, user-generated content, machine learning and autonomous vehicles. The company notes 53 supporting patents and an Emmy Award for Technology and Engineering, with deployment options spanning on-premises and major public clouds like AWS and Oracle Cloud.