Edge AI in Action: Promwad and Videology Bring Next-Gen Video Analytics to the SCAiLX Platform

At the XPONENTIAL exhibition in Houston, TX, USA, Videology will present a joint solution with Promwad: a real-time object detection system that runs directly on the SCAiLX embedded platform at the Edge — with no cloud, no servers, and minimal latency.

SCAiLX is a compact industrial platform by Videology that is designed to run AI analytics at the device level at the Edge. With a custom neural AI model developed by Promwad, it delivers fast and reliable object recognition in dynamic environments. All video processing happens onboard, making the system fully autonomous and secure.

This solution is ideal for a wide range of applications — from traffic monitoring to autonomous navigation, including situational awareness in UAVs, UGVs, and other uncrewed systems

Why Edge AI Is Critical

Cloud-based video analytics faces critical limits: high latency, overloaded networks, and security or compliance risks. Just one Full HD stream uses up to 10 Mbps — multiplied across dozens of cameras, bandwidth becomes a serious issue.

Edge AI solves these issues by processing data directly on the device — with low latency, no cloud dependency, and improved reliability. The SCAiLX platform with Promwad’s neural model is a robust solution for real-time, mission-critical applications.

Detecting Vehicles in Challenging Conditions

To demonstrate the capabilities of the SCAiLX platform, Videology commissioned Promwad to develop an AI solution for detecting vehicles from high angles — for example, for traffic control in logistics centers or at intersections in smart cities.

In practice, this meant that vehicles were displayed at a size of only 20–30 pixels, with additional challenges posed by factors such as variable lighting, lens distortion, and limited computing resources.

Examples of images from the training dataset

The Promwad team prepared a dataset based on publicly available drone videos. To ensure high accuracy in real-world conditions, the data was carefully adapted:

low-quality frames were removed;

object classes were merged;

additions were made to simulate real-world shooting conditions, including blur, shifts, and color distortion.

Identification of vehicles in illustrations from a dataset after training a neural model

YOLOv8 was chosen as the architecture, as it is one of the fastest and most accurate for edge analytics. After training, the Promwad team optimized the model for the SCAiLX platform and performed:

weight quantization (FP32 to INT8) for efficient execution time;

unsupported layers were adapted to the NPU architecture.

AI model performance results on SCAiLX: accuracy and speed

All processing is performed locally — no cloud storage is required. This allows the system to be used in areas with limited Internet access, reduces energy consumption, and increases overall reliability.

Scalable and Customizable for Real-World Use

The modular architecture of the solution allows it to be retrained or expanded for new classes of objects or operating conditions in accordance with It is already ready for use in industries where real-time response and autonomous operation are critical:

Eight application areas for edge video analytics with the SCAiLX AI camera — enabling real-time detection and automation across industries.

Ready to Build Your Own AI System on SCAiLX?

Promwad can help you:

Adapt your existing models for the SCAiLX platform.

Train AI models for your specific application.

Optimize software for on-device performance and low latency.

Support your project from PoC to full deployment.

Want to see the solution in action? Visit the Videology booth at XPONENTIAL.

Booth 3208 — Houston, TX

May 20–22, 2025

About Videology

Videology is a US-based manufacturer of embedded video cameras, offering high-performance modules for OEMs in automation, retail, security, healthcare, and other industries. Its SCAiLX platform brings real-time AI computing to the edge in compact, industry-ready form factors.