Embedded AI case: On-device AI video analysis for healthcare

What is Embedded AI?

It’s the combination of hardware and software systems that bring AI capabilities directly to compact devices.

These devices process various types of data locally, and among their perks are enhanced security, and the ability to empower any existing system, without rebuilding it.

But here’s the catch: even with modern and powerful processing units, Machine Learning models for such projects must be well optimized, and of a high quality.

Our latest case

We developed an AI assistant device that analyzes diagnostic video in near real-time. It uses object recognition technology to detect anomalies and provide AI-driven hints to clinicians — making diagnostics faster and smarter.

Object recognition requires significant computational power, so embedding this model into a portable scanner was a challenge. Both the Machine Learning and Embedded teams were involved:

• First, our ML team collaborated with healthcare professionals to develop the anomaly detection technology.

• Next, we optimized signal processing, frame trimming, and both pre- and post-processing stages to ensure smooth performance.

The result?

The AI model analyzes a video stream in near real-time, allowing for processing 30 frames per second. The model runs on the embedded controller, and it is processed locally. The result is a portable scanner that enhances diagnostic quality on the go.

The project brings lower costs, higher efficiency, and improved health outcomes for our client.

Embedded software for your business

WaveAccess's expertise encompasses the full development cycle of AI devices — from selecting the optimal hardware to crafting and integrating computer vision and machine learning algorithms to their optimization for resource-constrained environments.

We engineer smart devices that operate in real-time right at the edge, eliminating delays and cloud dependency. What we can create:

• Systems providing real-time assistance to medical professionals, for example, real-time capture and analysis of streaming video.

• Quality control, process automation, and complex equipment management – all locally, instantly, and without the need for cloud data transfers.

• Autonomous control systems to intelligent sensor data processing (AIoT).

Get in touch to explore how a custom embedded device can generate measurable results for your business.