Automated Installation Checks with AI and Computer Vision that Achieved 90% Accuracy
Learn how our AI Engineers helped a leading Dutch fibre-optic provider to save time and money on infrastructure installations and ensure product quality using cutting-edge computer vision quality inspection systems.
Leading European fibre optics provider can stay on top with an AI quality control system
One of the largest Dutch fibre-optic providers faced a critical operational challenge. Their installations were carried out by external contractors across homes, schools, and public institutions. To ensure quality, the client had a system to manually verify installation photos: contractors photographed the open box to confirm correct wiring and the closed box to confirm proper mounting. These photos were then reviewed by the client’s internal team before the installation could be approved.
While this system ensured quality, it was not scalable. With thousands of installations happening daily across multiple regions, and they were scaling their services to Germany, the manual verification process became a bottleneck. Approvals were slow, inconsistencies occasionally led to errors, and field teams sometimes had to return to correct mistakes. The client needed a way to automate these photo checks using AI so they could maintain reliable service, scale efficiently across regions, and continue providing high-quality installations without overloading internal teams.
👋 Hey You, What Are You Working On?
Share a few details about your challenge or idea, no matter how early or complex. Our expert team will reach out to offer our expertise in terms of feasibility, budget, and scope. Let’s shape a solution for you together!
The collaboration started with 2-day AI workshop to assess scope and ROI
The collaboration began with a two-day AI workshop. This workshop brought together stakeholders from the client and DAC.digital to align expectations, explore cost-effective solutions, and plan a quick implementation roadmap. The session allowed everyone to agree on the project’s objectives, define the technical approach, and ensure that the solution would be feasible at scale. We describe the workshop in detail in another case study, linked here: How GoConnectIT Used Workshops to Introduce AI to Their Process Without Slowing Field Operations.
Next, it was time to build the solution
After the workshops, the real challenges became clear. The client needed a team capable of handling sensitive massive data, creating a high-quality dataset, achieving accuracy, and deploying a scalable solution for thousands of installations.
The client’s primary expectation was clear: they needed an AI system that could reliably replicate or surpass the accuracy of their existing manual verification process. Initially, they estimated that achieving 60–70% accuracy would be sufficient, matching the performance of human reviewers. However, they also wanted the solution to be scalable and fast, capable of processing thousands of installation photos daily across multiple regions, including Germany, without introducing delays or errors. Beyond basic accuracy, the client expected the AI to handle the real-world variability of installation photos. Contractors took pictures in different lighting conditions, from varying angles, and with slight inconsistencies in framing. The AI had to correctly identify open and closed boxes, verify wiring and mounting quality, and provide results that could be trusted by internal teams. Essentially, the client wanted a system that was not only technically accurate but robust, reliable, and ready for real-world deployment at scale.
The scale and sensitivity of the data could have been a showstopper for many contractors
The client provided five terabytes of installation photos. It contained millions of images that needed to be transferred, processed, and annotated securely. Any misstep could have led to data loss, corruption, or unauthorized access. DAC.digital, however, found a way. We designed a fully secure workflow that allowed the team to access and manage the data efficiently, protecting sensitive information while keeping the project on track.
From this enormous pool, our team created a “golden dataset” of about 10,000 carefully annotated images, covering both open and closed installation boxes. This high-quality dataset became the foundation for training a reliable and accurate AI model.
AI doesn’t need to be a black box, so engineers added explainable AI for full visibility into how the system does visual inspections
Once the model was trained, we integrated Explainable AI (XAI) to show exactly how the AI made its decisions. XAI highlighted the regions of each image that the model considered when classifying an installation as correct or incorrect. This transparency allowed the client to understand the rules guiding the AI, verify its reliability, and extend the training dataset with new cases that the model initially missed.
By making the AI’s decision-making process clear, we built trust with the client and ensured that the system could be confidently used in real-world installations. Field teams could now rely on the AI to flag potential issues, speeding up approvals and maintaining consistent quality without manual oversight.
With the help of DAC.digital’s software team, a web app was created
The AI solution was delivered as a web application integrated with the client’s environment. Field teams can now upload a photo to the app, which automatically analyzes it to determine whether the installation is correct, identify the type of box, and confirm if it is open or closed. This real-time feedback significantly speeds up approvals and reduces the need for manual verification.
Need AI Engineers Who Are Experienced in Projects Like This?
Book a FREE intro call to learn more and see how we can help you get top talent from computer vision, machine learning and MLOps. Let’s talk!
The AI solution was made possible by a cross-functional team with AI, DevOps, and software skills
To deliver a reliable and scalable AI solution, the client relied on a dedicated team from DAC.digital. Two data analysts carefully curated and annotated a high-quality dataset from the client’s vast pool of installation photos, ensuring the model would learn from the most representative examples. Two machine learning engineers designed and trained the validation models, achieving the accuracy and reliability needed for real-world deployment.
The team was supported by MLOps and DevOps engineers who ensured seamless implementation and integration of the solution into the client’s environment. A project manager coordinated the workflow, keeping the project on track, while a single point of contact streamlined communication with the client. Together, this team acted as a trusted guide, turning a complex AI project into a practical, scalable solution that enabled the client to automate verification and confidently scale their operations.
Every Installation Is Right, Because AI Checks It
Ensure every setup is accurate, compliant, and efficient. DAC.digital builds custom AI models tailored to your hardware, installation procedures, and compliance requirements and integrates it into your workflow.
Technologies for Successful Photo Validation Solutions
- Python: the primary programming language
- PyTorch: Deep Learning
- Hydra: configuration file management
- GradCam: Explainable AI
- FastAPI: services
- OpenCV: image processing
- Numpy: matrix operations
Results: Fiber Optic Installation Verification is Fully Automated with AI
The AI solution was a success. We completed both stages: verifying indoor boxes and outdoor installations like fibre spools and excavations.
Our machine learning models were trained on a large dataset provided by the client. They achieved 90% accuracy, exceeding the client’s initial expectation of 60–70%. The client can now test the system in real-world conditions and compare it with manual photo checks.
Executives are preparing to roll out the solution across all installations. Once live, the process will be fully automated. It will save time, reduce costs, ensure consistent quality, and eliminate human errors.
Got a Project in Mind? Let’s Discuss It.
Contact us!
Send us an email: [email protected]
Read More of Our Case Studies.
Contact us!
Send us an email: [email protected]