Manufacturing systems across the United States are becoming more digitized, more connected, and increasingly data-driven, yet one of their most fundamental layers remains structurally uneven. Quality control still operates between two extremes that fail to reconcile in practice. On one side are manual measurement tools that are accessible but inconsistent, carrying error rates between 5% and 12% and introducing variability that compounds across production volumes. On the other are automated metrology systems that deliver high precision but often cost upwards of $50,000 and require trained specialists to operate. This divide is not a marginal inefficiency. It contributes to an estimated $10 billion annually in scrap, rework, and downtime, creating a systemic productivity gap that persists despite broader advances in manufacturing technology.
Qiushi Wang, Founder-CEO of Labro Inc., and a judge for the SARC and ESP journals works directly within this gap. With a background in computer vision, biomedical engineering, and precision instrumentation, his work focuses on making measurement systems deployable under real production constraints rather than ideal conditions. His work spans both physical measurement systems and algorithmic pipelines, positioning him at the intersection of inspection, data translation, and manufacturing digitization.
“Most manufacturing problems are not about whether a solution exists,” Wang explains. “They are about whether that solution can actually be used inside the workflow without slowing everything down or requiring specialized support.”
This distinction defines the challenge at hand. The issue is not the absence of precision, but its uneven distribution. As manufacturing systems move toward digital continuity and integrated data flows, the inability to apply consistent, reliable measurement across environments becomes a limiting factor that no amount of downstream automation can fully resolve.
Why Precision Remains Structurally Out of Reach
The prevailing assumption in manufacturing modernization is that automation and AI adoption will gradually close operational gaps. In reality, the constraint is more foundational. Precision itself remains inaccessible to a large portion of manufacturers, particularly those operating without the capital or staffing required for advanced metrology systems. Manual tools persist because they are deployable, not because they are sufficient. Their limitations become visible at scale, where small measurement errors translate into rework cycles, production delays, and variability that erodes consistency across batches.
At the same time, high-end metrology systems concentrate precision within controlled environments rather than distributing it across workflows. Their cost and operational complexity limit where they can be deployed, effectively turning precision into a centralized capability instead of a pervasive one. This imbalance becomes more pronounced as workforce constraints tighten, making reliance on specialist operators increasingly difficult to sustain.
Wang’s work on an image-based instant vision measurement system addresses this imbalance by removing the conditions that restrict deployment. The system uses a calibrated optical setup combined with contour extraction algorithms to achieve sub-0.1 mm dimensional accuracy, while reducing system cost to under $2,000 per unit. More importantly, it is designed for operator-level usability, allowing measurement to be integrated directly into production workflows without requiring specialized training. In testing environments, this approach has enabled inspection processes to run up to 10 times faster, while eliminating dependence on touch-probe devices and reducing operator-induced variability.
“In many environments, measurement is treated as a separate step because it interrupts production,” Wang notes. “If you can bring it into the workflow without friction, you remove the reason people skip it in the first place.”
The impact is not only technical but structural. By lowering the cost and complexity of precision, the system allows manufacturers to move away from reactive inspection toward continuous, embedded quality control, aligning measurement with the pace of production rather than placing it outside of it.
From Vision as Detection to Vision as Measurement
Much of the current adoption of computer vision in manufacturing is centered on defect detection. Systems are trained to identify anomalies and classify nonconforming parts, acting as a downstream filter for production errors. While effective, this approach addresses only part of the quality problem. Detection identifies failure after it has occurred, but it does not provide the information required to correct the process that produced the deviation.
The more consequential shift occurs when vision systems move from detection to measurement. When a system can extract contours, quantify dimensions, and generate structured outputs, it becomes part of the process itself rather than a checkpoint at the end. This transition changes how quality is managed, shifting from rejection-based workflows to process-level control.
Wang’s system reflects this shift by combining calibrated imaging with edge detection and geometric reconstruction to produce CAD-aligned measurement outputs. These outputs allow inspection data to feed directly into engineering workflows, enabling repeatability and process correction rather than isolated defect identification. His publicly documented work, including his HackerNoon article, “Building a DIY Image-Based Part Inspection System: No Expensive Hardware Needed”, emphasizes that accessible measurement is the missing layer in current vision deployments.
His perspective is further informed by his role as a judge at the World Conference on Science Engineering and Technology, where he evaluated systems that claim to bridge research and production. That experience exposes a consistent pattern. Many vision systems perform well in controlled demonstrations but fail to deliver stable, repeatable measurements under production variability.
“The difference shows up when conditions are not ideal,” Wang explains. “Production environments are noisy, inconsistent, and time-constrained. If a system cannot handle that, it is not solving the real problem.”
This distinction elevates vision from a support tool to a component of quality infrastructure, enabling manufacturers to influence processes in real time rather than reacting to outcomes after the fact.
When Engineering Drawings Break Digital Continuity
Even as inspection becomes more accessible, manufacturing digitization encounters another constraint upstream. Engineering drawings, particularly those stored as raster images or legacy blueprints, remain difficult to convert into structured geometry. This creates a break in the digital thread, where design intent cannot be seamlessly translated into manufacturing and inspection workflows.
This issue is less visible than shop-floor inefficiencies but equally limiting. When geometry must be manually interpreted, CAD reconstruction slows, inspection planning becomes fragmented, and reverse engineering workflows lose consistency. The result is a disconnect between design and execution that undermines the effectiveness of otherwise advanced production systems.
Wang’s geometry-guided computer vision pipeline addresses this bottleneck by focusing on automated extraction of geometric features from engineering drawings. Using a combination of four-point geometric sampling, multi-stage filtering, clustering, and machine-learning-based reranking, the system generates and refines candidate geometries directly from image inputs. Compared with classical methods such as Hough Transform and RANSAC, it improves F1 score from 0.23 to 0.80 while reducing reconstruction time by over 90 percent.
These improvements translate directly into operational impact. Tasks that previously required manual interpretation and repeated validation can be processed at scale with consistency, enabling downstream systems to operate on structured inputs rather than approximations. The work, supported through his role as an invited Judge at the EuroHaptics 2026, highlights a broader reality. Manufacturing digitization depends not only on automation but on the ability to make existing data usable.
“If you cannot turn the drawing into something the system understands, you are not actually running a digital process,” Wang says. “You are still relying on interpretation.”
Accessible Precision Will Define Competitiveness
Manufacturing modernization is often framed around automation, AI adoption, and production scale. These factors are visible, but they do not fully determine system performance. The more fundamental question is whether precision can be deployed consistently across the environments where production decisions are made.
The current landscape suggests that this remains unresolved. A large portion of manufacturers continue to operate between manual tools that introduce variability and high-cost systems that limit scalability. The resulting inefficiencies are structural, shaping cost, throughput, and reliability across the production lifecycle.
Wang’s work connects two layers of this problem. His image-based measurement system addresses how precision is applied on the shop floor, while his geometry-guided extraction pipeline addresses how design data is translated upstream. Together, they align measurement and data in a way that supports continuous improvement rather than isolated optimization. “Precision should not depend on where you are in the system,” Wang concludes. “If it is critical, it should be available everywhere the work is happening.”
As manufacturing continues to evolve, that principle becomes decisive. The ability to distribute precision across workflows, rather than confine it to specialized environments, will define how effectively manufacturers can operate in a fully integrated, data-driven production model.
