Frequently Asked Questions

We offer three models tailored to different needs. The OV10i is our entry-level option with classification capability, built-in lens, NVIDIA Xavier NX GPU, and 1.6MP resolution at $4,450. The OV20i adds segmentation, features a C-mount for lens flexibility, NVIDIA Orin NX GPU, and 8.4MP resolution at approximately $9,450. The OV80i is our premium offering with best-in-class classifier, segmenter, and OCR capabilities at $13,450.

No external compute is required. All three models have onboard GPU processing integrated directly into the camera, so inspections run completely standalone at the edge without needing servers, cloud connections, or additional hardware.

For OV80, we offer standard C-mount lenses plus specialized options for challenging applications. Telecentric lenses eliminate perspective distortion for flat-field accuracy, while our photometric control system captures four images with different directional lighting and combines them to handle reflective surfaces. LED strobe mode reduces reflections and motion blur on fast-moving parts.

The OV10i and OV20i can detect defects as small as 0.5% of the field of view, while the OV80i achieves sub-0.1% FOV sensitivity for catching micro-defects and critical quality issues that other systems might miss.

The OV10i stores approximately 300,000 images, while the OV20i and OV80i each store around 75,000 images locally before transfer to external systems.

Yes. We support 360-degree inspection through strategic use of specialty mirrors in your inspection setup, allowing complete part surface coverage with a single camera. This has been proven in wire braiding applications across multiple global facilities.

Typically just 5–20 images depending on application complexity. Training time ranges from 15 minutes for simple classifications to 4 hours for complex segmentation tasks with multiple defect types. Most customers have a production-ready model the same day.

Our models consistently deliver 97–100% accuracy across diverse applications including electronics, pharmaceuticals, automotive, and wire/cabling. We've achieved 100% accuracy with zero escapes and less than 1% overkill across 1,000+ deployed cameras for a customer.

Classification assigns a label to a region of interest (pass/fail, part type A vs. B). Segmentation provides pixel-level detection of where defects exist, giving precise defect location, size, and shape information. Use classification for simple pass/fail and segmentation for mapping exactly where defects are.

Our Gen-AI Defect Creator can synthetically generate defect variations to expand your dataset. We've also successfully trained models on marker-drawn 'fake' defects that generalized to catching real production defects. Built-in data augmentation (brightness, contrast, rotation, noise, motion blur) further extends limited training data.

Haystack is a secondary AI that continuously monitors inspection quality, flagging anomalies and edge cases for human review. It helps identify when lighting drifts, parts vary unexpectedly, or your model needs retraining, enabling continuous improvement and catching issues before they impact production.

Yes to all three. Photometric control with directional lighting handles reflections, LED strobe mode captures moving parts without blur, and our pre-trained vision transformer models are specifically optimized for low-contrast defect detection on challenging materials.

We support EthernetIP, Profinet, Modbus, HTTP/HTTPS, TCP/IP, FTP, SFTP, API, MQTT, WebSocket, OPC UA, and SMB. This compatibility covers virtually all major industrial automation ecosystems.

We integrate with Allen-Bradley, Siemens, and other major PLC brands through standard industrial protocols. Our application engineers handle the configuration to ensure seamless communication between the camera and your production line controls.

We offer five trigger modes: PLC trigger (automation-controlled), hardware trigger (physical sensors), interval-based (fixed time between inspections), alignment trigger (triggered by part position detection), and API trigger (external software control).

No. We provide a browser-based UI accessible directly from your network with no software downloads required. For advanced users, we offer Node-RED for custom logic and our new Agentic AI flow builder that uses natural language to configure integrations.

Yes. With appropriate C-mount lenses, our cameras can detect defects from 7+ feet away. At a customer's pipe facility, a single OV80i detected 7 defect types on metal pipes from 7 feet away with 97–98% accuracy.

Our vision transformer AI is pre-trained on millions of industrial images, so it understands manufacturing defect patterns before it ever sees your parts. Combined with photometric control, telecentric lenses, and advanced segmentation, we routinely solve applications that legacy vision systems couldn't handle—including black-on-black scratches, micro-cracks, and shiny surface inspections.

We deployed 1,000+ cameras for a customer across 6 countries (China, Vietnam, Japan, Taiwan, Philippines, Malaysia) in 12 months, achieving zero escapes and saving over $25M annually. This deployment included 700+ devices at a single facility.

It synthetically generates realistic defect images to augment limited training data, enabling up to 10x faster deployments. This is particularly valuable for rare defects or safety-critical applications where collecting hundreds of real defect samples isn't practical.

Absolutely. We've deployed single-camera solutions detecting 7+ defect types in one recipe. Our PCBA soldering application detects 7 distinct defect types simultaneously. The pipe inspection at a customer site handles 7 inspection types with one camera.

No. Our pricing is all-inclusive hardware with no recurring software fees. You own the camera outright—no subscriptions, no annual licensing, no per-image costs. Free software updates are included.

Most customers see payback in 2–5 weeks depending on labor costs and defect escape rates. Our fastest documented payback was 5 days for a robotic part placement application. A pharmaceutical tablet application achieved 3-week payback.

Yes. We offer a 30-day money-back guarantee. Deploy the camera, run it through your toughest scenario, and if it doesn't deliver, return it for a full refund. We also provide free onsite training with every deployment.

Every customer receives a dedicated application engineer, free onsite training, and ongoing access to our technical team throughout deployment and beyond. We staff engineers specifically for customer success.

Training typically takes 1–2 days and is tailored to your specific application and team. Our browser-based interface is intuitive enough that engineers can master the software in under 1 hour for basic operations.

Firmware updates are managed remotely and applied on your schedule—not automatically pushed. This gives you control over testing and validation before deployment, which is essential for regulated industries.

We offer OV Vault and OV Fleet—centralized management platforms for fleets at scale. You can monitor performance, update firmware, aggregate data, and manage models across hundreds of cameras from a single dashboard with role-based access.

Yes. Train a model once, deploy it across your fleet. Our fleet management tools let you version models, roll out updates gradually, and track performance across all cameras. Wire braiding customers have successfully copy-pasted models across global facilities with no retraining.

After your first deployment, subsequent cameras integrate faster because you've established processes and trained staff. One customer scaled from initial pilots to 1,000+ cameras in 12 months. OV Fleet management automates many tasks that would otherwise be manual at 5+ cameras.