Syntiant to Unveil Vision Transformer Security Solution at ISC West 2025

April 02, 2025 at 07:03 AM EDT

IRVINE, Calif., April 02, 2025 (GLOBE NEWSWIRE) -- Syntiant Corp., the recognized leader in low-power edge AI deployment, today announced it will demonstrate its multimodal vision transformer (ViT) at ISC West 2025 in Las Vegas, April 2-4. Deployed on an Ambarella CV75M system on chip (SoC), Syntiant’s real-time edge AI security solution significantly enhances security camera performance with smarter, faster and more efficient monitoring, reducing latency and costs while enhancing privacy.

“Transformers have revolutionized language models by enabling highly efficient, context-aware processing, improving accuracy and performance. Syntiant is applying this same transformer-based architecture to computer vision, significantly enhancing real-time image processing, making it a game-changer for security applications,” said Greg Coladonato, product line manager for AI models at Syntiant. “Our AI vision security solution on Ambarella hardware sets a new benchmark for low-power, real-time performance, revolutionizing how security teams track, search and analyze data across multiple cameras, while providing customers with unprecedented value with each added sensor.”

Syntiant’s new security solution is ideal for a wide range of applications, including real-time public safety, traffic monitoring, smart home management and automated system management across urban, residential and enterprise environments.

“Tomorrow’s security use cases will depend upon advanced capabilities offered by vision transformers running at the edge,” said Shay Kamin Braun, director of marketing at Ambarella. “Our demonstration of Syntiant’s vision transformer optimized for Ambarella’s CV75M is the next step in our long-standing partnership with Syntiant. We look forward to continuing our work with them productizing and deploying state-of-the-art AI models.”

Next-Level Visual Intelligence with Vision Transformers

Syntiant’s vision transformer is advancing computer vision by processing visual data as sequences, capturing complex relationships in imagery, and mapping text and images to a common latent space. Unlike traditional CNN-based models that require tens of thousands of images to train and deploy production-grade models, transformer-based models are more generalized and enable zero-shot classification, virtually eliminating the need for additional data.

Syntiant has pre-tuned its edge-based ViT model for security use cases, delivering over 10% greater accuracy than open-source vision transformer models, while reducing training data requirements by 1000x relative to traditional CNNs. Paired with Ambarella’s CV75M, which features a CVflow^® NPU engine for hardware-accelerated transformer processing, Syntiant’s model achieves a 30x speedup over ARM CPU-based implementations, enabling faster, more flexible deployments into production.

Syntiant’s ViT also integrates seamlessly with its other machine learning models, such as its Small Language Model Assistant (SLMA), which provides a natural voice interface. Users can issue commands like “find people wearing yellow jackets” without requiring prior model training.

Key features of Syntiant’s ViT-based solution include:

Low-power edge-based AI processing: Inference is performed on-device, reducing latency and enhancing privacy, while balancing high performance with energy efficiency.
Image-text similarity: Instantly interpret ad-hoc text or visual descriptions (e.g., “person in the yellow jacket”) without needing to be pre-trained on specific search terms.
Zero-shot classification: Identify previously unseen objects using generalized visual-linguistic models, expanding recognition capabilities without additional model training.
Cross-camera tracking: Simultaneous recording of the same subject matter or individual across multiple devices, eliminating redundancies and improving situational awareness.
Seamless model integration: Works with Syntiant’s existing suite of ML models, including detectors, analytics, wake words and voice commands.

Revolutionizing Security Applications with Edge AI

Syntiant’s hardware-agnostic AI models offer industry-leading inference speed and minimized memory footprints, making them ideal for a range of applications, including:

Detection: Identify and localize objects (i.e., people, vehicles) in an image.
Classification: Provide detailed categorization of objects (e.g., bicycles, trucks).
Segmentation: Background removal and contextual scene analysis.
Acoustic Event: Detect audio-based events including glass break, smoke alarms and carbon monoxide alarms.

Ambarella’s CV75 and CV75M families of integrated circuits combine 4KP30+ image processing, video encoding/decoding and CVflow^® computer vision capabilities in a compact, low-power design. Built on advanced 5 nm process technology, the CV75 delivers ultra-high-definition imaging with optimized frame rates, high dynamic range (HDR) support and enhanced low-light performance. The SoC’s built-in CNN toolkit simplifies neural network deployment from popular frameworks such as Caffe, TensorFlow, PyTorch and ONNX.

ISC West 2025

The new security solution will be demoed at Syntiant’s suite and Ambarella’s booth (Veronese Ballroom #2403 at The Venetian Convention & Expo Center) from April 2-4 at ISC West 2025. Contact sales@syntiant.com for more information or to schedule a demonstration.

About Syntiant

Founded in 2017 and headquartered in Irvine, Calif., Syntiant^® is Making Edge AI a Reality™ by delivering highly efficient processor, sensor and software solutions. With more than 100 million purpose-built silicon and ML models deployed, along with billions of MEMS microphones and sensors, Syntiant’s technology is powering edge AI applications for speech, audio, sensor and vision processing worldwide. From earbuds to automobiles, the company’s turnkey solutions enable advanced edge AI capabilities across diverse consumer and industrial use cases. More information on the company can be found by visiting www.syntiant.com or by following Syntiant on Twitter @Syntiantcorp or LinkedIn.

Media Contact:

George Medici
PondelWilkinson Inc.
gmedici@pondel.com
310-279-5968