Empowering Security with AI-Driven Intrusion Detection: A Crow-Free Zone at a Japanese Factory

A Japanese factory was facing an unusual but persistent problem: a large population of crows had taken up residence on its rooftop. Beyond the visual impact, the birds were creating hygiene concerns through their droppings — contaminating surfaces, posing health risks to workers, and damaging the factory's image. The factory needed an effective way to deter the crows continuously and automatically, without causing them any harm and without assigning personnel to monitor the roof around the clock. The answer was an AI Vision intrusion detection system built around the Aetina AIP-SQ67 edge AI computing platform.

The Challenge: Persistent, Harmful, and Hard to Monitor

Crow infestations in industrial environments are more than a nuisance. Droppings are corrosive to rooftop materials and equipment, carry pathogens that present hygiene risks, and can cause regulatory compliance issues in food-adjacent or pharmaceutical manufacturing facilities. Traditional deterrence methods — physical spikes, netting, manual patrols — either require costly installation, ongoing maintenance, or continuous human labor. None of them respond dynamically to crow activity in real time.

What the factory required was a system that could detect the presence of crows automatically and immediately trigger a deterrence response — without human intervention, 24 hours a day. Crucially, the solution had to be humane: no physical traps, no chemicals, no harm to the birds. The detection range also had to be long enough to cover the full rooftop area, with the ability to connect to the factory's existing camera and NVR infrastructure without a full hardware replacement.

The Solution: Aetina AIP-SQ67 + IronYun Vaidio DIY

The deployed solution combines the Aetina AIP-SQ67 — a high-performance x86 MegaEdge AI inference platform — with IronYun's Vaidio DIY software, an eco-partner AI vision platform designed for fast object recognition training and deployment. Together, they form an end-to-end AI intrusion detection system that required no changes to the factory's existing ONVIF-compatible IP cameras or NVR systems.

Vaidio DIY's low-code training workflow allowed the solution team to train a custom AI model specifically to recognize crows — their shape, movement patterns, and roosting behaviour — in a short amount of time, without extensive machine learning expertise. Once trained and deployed on the AIP-SQ67, the model runs inference continuously on the live camera feeds. When a crow is detected within the monitored zone, the system immediately triggers the connected 3rd-party speakers to emit a loud deterrent sound, dispersing the birds before they settle.

How the System Works

• Camera feeds from existing ONVIF-compatible IP cameras are streamed to the AIP-SQ67 over the factory network — no new camera hardware required.
• AI inference runs locally on the AIP-SQ67, processing multiple simultaneous camera channels in real time using the onboard Intel Core processor and optional MXM AI accelerator module.
• Object detection is performed by a custom-trained model built with IronYun Vaidio DIY, specifically tuned to identify crows at distances of 50 to 70 meters.
• Deterrence trigger is issued automatically via integration with 3rd-party speakers connected to the system — loud sounds are emitted immediately upon detection, dispersing the crows humanely.
• Adjustable parameters allow the sensitivity, detection zones, and deterrence response to be fine-tuned to the specific geometry and requirements of each rooftop area.

About the Aetina AIP-SQ67

The AIP-SQ67 is Aetina's MegaEdge expandable AI inference platform designed for vision-based AI applications that demand high multi-channel throughput, rich I/O connectivity, and the flexibility to add GPU or AI accelerator modules as application requirements grow. Its Intel 12th/13th Gen Core processor foundation, combined with a PCIe Gen4 x16 MXM slot supporting NVIDIA RTX and Hailo-8 accelerator modules, makes it equally capable as a standalone inference engine and as an upgradeable platform for demanding workloads.

For the crow detection application specifically, the AIP-SQ67's five 2.5GbE LAN ports and five DisplayPort outputs allow simultaneous connection and display of multiple camera streams, while its built-in out-of-band (OOB) remote management module — accessible via the EdgeEye platform — enables the factory's IT team to monitor system health, push configuration updates, and perform remote recovery without physically accessing the rooftop installation.

Key Technical Specifications

Broader Applicability: Beyond Crow Detection

While this deployment addresses a very specific use case, the underlying solution architecture — AIP-SQ67 + Vaidio DIY + existing IP camera infrastructure — is readily applicable to a wide range of security and intrusion detection scenarios. The same platform can be retrained to detect unauthorized personnel in restricted zones, monitor perimeter fences for breaches, identify specific animals or objects in agricultural or logistics environments, or trigger automated responses to any defined visual event. The ease of retraining with Vaidio DIY means that adapting the solution to a new detection target requires days rather than months of development effort.

The AIP-SQ67 is also used in other demanding AI Vision applications including Automated Optical Inspection (AOI) in manufacturing, 3D virtual fence systems for workplace safety, multi-channel video analytics for smart city surveillance, and 4K surgical imaging systems in medical environments — demonstrating the breadth of the platform's capabilities across industries.