Training AI model on video data processed by Beamr's content-adaptive technology made the model more resilient to compression, by lowering depth estimation error on safety-critical road users, including pedestrians and motorcyclists, by 30.7%

Herzliya, Israel, May 06, 2026 (GLOBE NEWSWIRE) -- Beamr Imaging Ltd. (NASDAQ:BMR), a leader in video optimization technology and solutions, released research demonstrating that machine vision models fine-tuned on video compressed by Beamr's patented Content-Adaptive Bitrate (CABR) technology are more resilient than models trained on uncompressed data, while reducing the video data volumes that machine vision development depends on.

Machine vision teams handling petabyte-scale video data for autonomous vehicles (AV) and other video AI applications typically consider compression as a process for managing this scale. The findings reframe adaptive compression as an asset that strengthens AI model resilience, with the advantages of reducing storage and networking costs and infrastructure. This research extends Beamr's ML-Safe benchmarks, validating a potential performance asset for AI models trained across machine vision applications.

The research evaluated Depth Anything V2, a state-of-the-art monocular depth estimation model. The model was fine-tuned on AV video data compressed with Beamr's technology that delivered 35.2% file-size reduction relative to baseline compression. The fine-tuned model demonstrated 30.7% reduction in depth estimation error on vulnerable road users, including pedestrians and motorcyclists, and 16.0% aggregate reduction across all object classes. Full methodology and results are available in the blog post.

"This research shows that compressed video data can produce models that are more robust, not less," said Dani Megrelishvili, Beamr CPO. "That points to a different role for compression in our customers' pipelines, from a cost they tolerate to a tool they deploy."

"Machine vision teams have faced a structural trade-off: compress video data to manage scale, or face the escalating costs and infrastructure challenges of running AI models without compression," said Ronen Nissim, ML Lead at Beamr. "Our research suggests this trade-off is more flexible than the industry may have assumed. By using compressed footage as augmentation during fine-tuning, we produced a model that performed better on the validation set than the equivalent model trained on uncompressed data."

Beamr's ML-safe benchmarks have previously validated content-adaptive compression across the AV development pipeline. The benchmarks demonstrated up to 50% file size reduction while preserving object detection accuracy at mean average precision of 0.96, with high fidelity across detection, localization, and confidence consistency. Subsequent testing for captioning workflows in world foundation model pipelines showed 41%–57% file size reduction with no measurable impact on the pipeline outputs.

To run Beamr's compression on your own data, visit beamr.com/autonomous

About Beamr

Beamr (NASDAQ:BMR) is a world leader in content-adaptive video compression, trusted by top media companies including Netflix and Paramount. Beamr's perceptual optimization technology (CABR) is backed by 53 patents and a winner of Emmy® Award for Technology and Engineering. The innovative technology reduces video file sizes by up to 50% while preserving quality and enabling AI-powered enhancements.

Beamr powers efficient video workflows across high-growth markets, such as media and entertainment, user-generated content, machine learning, and autonomous vehicles. Its flexible deployment options include on-premises, private or public cloud, with convenient availability for Amazon Web Services (AWS) and Oracle Cloud Infrastructure (OCI) customers.

For more details, please visit www.beamr.com or the investors' website www.investors.beamr.com

Forward-Looking Statements

This press release contains "forward-looking statements" that are subject to substantial risks and uncertainties. Forward-looking statements in this communication may include, among other things, statements about Beamr's strategic and business plans, technology, relationships, objectives and expectations for its business, the impact of trends on and interest in its business, intellectual property or product and its future results, operations and financial performance and condition. All statements, other than statements of historical fact, contained in this press release are forward-looking statements. Forward-looking statements contained in this press release may be identified by the use of words such as "anticipate," "believe," "contemplate," "could," "estimate," "expect," "intend," "seek," "may," "might," "plan," "potential," "predict," "project," "target," "aim," "should," "will" "would," or the negative of these words or other similar expressions, although not all forward-looking statements contain these words. Forward-looking statements are based on the Company's current expectations and are subject to inherent uncertainties, risks and assumptions that are difficult to predict. Further, certain forward-looking statements are based on assumptions as to future events that may not prove to be accurate. For a more detailed description of the risks and uncertainties affecting the Company, reference is made to the Company's reports filed from time to time with the Securities and Exchange Commission ("SEC"), including, but not limited to, the risks detailed in the Company's annual report filed with the SEC on February 26, 2026 and in subsequent filings with the SEC. Forward-looking statements contained in this announcement are made as of the date hereof and the Company undertakes no duty to update such information except as required under applicable law.

Investor Contact:

investorrelations@beamr.com

Beamr Research Validates Patented CABR Technology as an AI Training Asset