ACI OS Replaces Manual Inspection, Reducing Capsule Production Labor Costs

DaoAI ACI OS (featuring vision foundation model-based feature recognition, 5-minute 0-code automatic programming with one good sample, APDT positive/few-shot learning from 1–20 good samples, semantic false positive filtering, and SDK/API/Docker support for 100% on-premise deployment) has introduced an automated visual inspection solution, reducing the manual inspection workload for capsule appearance defects by 85% and overall inspection costs by over 40% for a medium-sized pharmaceutical company, significantly enhancing production line automation and economic benefits.

In the pharmaceutical industry, particularly in capsule production, strict control over product appearance quality is paramount. Even minor defects can compromise drug efficacy, safety, and brand reputation. Traditionally, the inspection of capsule appearance defects heavily relies on manual visual inspection, especially in small and medium-sized pharmaceutical companies where the high cost of automated equipment makes manual inspection the predominant method. While flexible, this approach faces challenges such as low inspection efficiency, fluctuating missed detection rates, long personnel training cycles, continuously rising labor costs, and high labor intensity. With increasing market demands for drug quality and escalating labor costs, seeking an efficient, stable, and economical automated inspection solution to replace or significantly reduce manual inspection has become an urgent need for pharmaceutical enterprises.

Pain Points: Why This Hurdle Is So Difficult to Overcome

Manual visual inspection of capsule appearance defects, in small and medium-sized pharmaceutical companies, typically accounts for 15%–25% of total production labor costs. For a capsule production line with an annual output value of several hundred million RMB, the annual labor cost for manual inspection alone could reach millions of RMB. Furthermore, the missed detection rate for manual inspection usually fluctuates between 0.5%–1.5% and is susceptible to factors such as inspector fatigue, mood, and experience, leading to quality variations between batches. Simultaneously, a skilled inspector requires 3–6 months of professional training to be qualified, and high turnover rates further increase management burden and training costs for enterprises. Traditional rule-based vision systems often struggle to achieve ideal results when dealing with the unique reflective and semi-transparent materials of capsules and varied defect morphologies, leading to high false positive rates. This necessitates extensive manual re-inspection time, further driving up operational costs. These factors collectively create a dual dilemma for enterprises in quality control and labor costs.

The difficulty in capsule defect detection lies in its diversity and subtlety. For example, capsule surfaces may exhibit various types of defects such as color spots, dents, damage, air bubbles, adhesions, and foreign matter, with sizes ranging from tens of micrometers to several millimeters, and diverse morphologies. The optical properties of capsule materials (e.g., high reflectivity, semi-transparency) make defects difficult to image stably under conventional lighting, easily generating artifacts or masking real defects. Moreover, on high-speed production lines, the inspection cycle is fast, leaving minimal time for the inspection system to process images. Traditional 2D vision systems struggle to effectively capture the 3D morphological features of capsules, while 3D stereo cameras, though capable of acquiring richer morphological information, present challenges in practical pharmaceutical line applications due to large data processing volumes, complex algorithms, high costs, and sensitivity to ambient light and vibration, making integration and maintenance difficult. DaoAI ACI OS addresses these challenges by combining vision foundation models and few-shot learning capabilities, providing an efficient solution for the pharmaceutical industry.

Technical Principles

The DaoAI ACI OS effectively replaces manual visual inspection and reduces labor costs primarily due to its vision foundation model-based feature recognition capabilities and the APDT positive/few-shot learning mechanism. The system first utilizes pre-trained vision foundation models to perform deep feature extraction and semantic understanding of capsule images. Unlike traditional rule-based AOI, which relies on manually set thresholds and feature extractors, DaoAI's model autonomously learns and identifies complex defect patterns, including subtle color differences, dents, or foreign matter that are difficult for the human eye to perceive. This enables it to effectively distinguish real defects from imaging artifacts when dealing with the unique highly reflective and semi-transparent materials of capsules, reducing the false positive rate by over 70% and significantly decreasing the need for manual re-inspection.

Furthermore, the APDT (Adaptive Positive/Few-shot Defect Training) technology introduced by DaoAI ACI OS allows users to complete model training with just 1-20 good samples, achieving 5-minute 0-code automatic programming. This means that even with frequent changeovers or new product launches, there is no need for professional vision engineers to perform tedious parameter adjustments, significantly shortening changeover downtime and reducing reliance on technical personnel. Compared to traditional manual inspection, it eliminates missed detections and misjudgments caused by human eye fatigue and subjective judgment; compared to traditional rule-based AOI, its intelligent semantic false positive filtering function effectively distinguishes normal texture variations of the product itself from actual defects, avoiding extensive rule maintenance and high costs associated with false positive re-inspection. Additionally, DaoAI ACI OS supports 100% on-premise private deployment via SDK/API/Docker, ensuring sensitive production data remains in-house, meeting the strict requirements of the pharmaceutical industry for data security and compliance.

Typical Application Scenarios

• **Capsule Surface Color Spots and Stains Detection:** Targeting uneven coloring, dark/light spots, or stains acquired during production on the capsule surface. The DaoAI ACI OS, leveraging its powerful color perception and texture analysis capabilities, can precisely identify areas with color differences exceeding a specific threshold from the baseline, distinguishing between material batch variations and actual defects to avoid false positives.
• **Capsule Breakage and Dent Detection:** Identifying cracks, holes, or localized deformations that may occur in the capsule shell during manufacturing, filling, or packaging. Traditional 2D vision struggles to capture subtle morphological changes effectively, while the DaoAI software, combined with high-resolution imaging, analyzes pixel gradients and edge features to accurately locate micrometer-scale breakages and dents.
• **Capsule Adhesion and Foreign Matter Detection:** Detecting whether capsules are stuck together or if foreign matter such as fibers or dust is adhering to the surface. The DaoAI ACI OS performs fine analysis of capsule contours and surface features to identify abnormal connection structures or non-product attachments, ensuring product purity.
• **Capsule Size and Integrity Detection:** Ensuring capsule length and diameter meet standards, and checking for integrity issues such as chipped corners or deformations. The system can perform high-precision geometric measurements and real-time inspection at high speeds, rejecting non-conforming products to ensure product consistency.

Case Study

A medium-sized pharmaceutical company, primarily manufacturing various specifications of hard and soft capsules, previously relied heavily on manual visual inspection for its production lines. As product variety increased and production scale expanded, the labor costs and management difficulties associated with manual inspection became increasingly prominent. Annual labor costs for manual inspection alone reached up to 3 million RMB, and during peak seasons, the company faced challenges in recruitment and high personnel turnover rates. Moreover, the average missed detection rate for manual inspection was around 0.8%, occasionally leading to customer complaints and impacting brand reputation. The company decided to introduce the DaoAI ACI OS, aiming to replace manual inspection and optimize its labor structure. After initial surveys and production line modifications, the company deployed the DaoAI ACI OS on a hard capsule production line, integrating it with existing camera equipment. After system go-live, model training was completed with just 10 good sample images using APDT, and it was quickly put into operation.

“After the DaoAI ACI OS went live, our production line's manual inspection workload was reduced by 85%, saving over 2 million RMB in labor costs annually. Product quality stability also significantly improved, with the missed detection rate dropping to <0.2%.”

Before deployment, this production line required 8-10 inspectors working in three shifts. After deployment, only 1-2 operators are needed for equipment monitoring and a small amount of re-inspection, reducing the manual inspection workload by 85%. The system's false positive rate also dropped from 5%–8% with traditional rule-based AOI to <1%, greatly reducing the workload for manual re-inspection. Overall inspection efficiency increased by 3 times, with production rhythm improving from 12,000 capsules per minute to over 15,000 capsules per minute. More importantly, the DaoAI ACI OS consistently controlled the missed detection rate to <0.2%, significantly lower than manual inspection levels, effectively improving product quality uniformity and customer satisfaction. Through this deployment, the enterprise not only solved the problem of high labor costs but also gained valuable experience for the intelligent upgrade of other production lines.

DaoAI Solutions and Products

DaoAI provides the pharmaceutical industry with an ACI OS, centered on its powerful vision foundation models and APDT few-shot learning capabilities, achieving intelligent and automated inspection of capsule appearance defects. In practical implementation, we first integrate the software system with existing or new vision hardware (industrial cameras, lighting, controllers) via SDK/API/Docker, ensuring 100% on-premise private deployment and secure data control. During the modeling phase, users only need to provide 1–20 good samples, and the DaoAI ACI OS can automatically complete model training within 5 minutes, generating efficient inspection models without any code writing. For multi-variety, small-batch capsule production, this 0-code rapid changeover capability significantly reduces production line downtime. The system's semantic false positive filtering function effectively distinguishes normal morphological features of the product itself from actual defects, avoiding false positives and minimizing manual re-inspection.

In addition to the core ACI OS, DaoAI also offers the DaoAI World Model as a unified foundation, enabling semantic understanding, cross-scene generalization, and continuous learning from production line feedback, further enhancing inspection intelligence and adaptability. If customers have higher requirements for 3D morphological defects, they can consider integrating DaoAI 2D/3D ACI equipment, utilizing self-developed 3D cameras for 3D morphological reconstruction to detect micrometer-scale dents or hidden defects. Overall, the DaoAI ACI OS consistently reduces the missed detection rate for capsule appearance defects to <0.2%, while the false positive rate is reduced by over 70%, significantly cutting down the manual re-inspection workload. Ultimately, customer annual labor costs are reduced by over 40%, product quality is significantly improved, and market competitiveness is enhanced. Our solution represents not only a technological upgrade but also an optimization of production models and cost structures, providing solid support for the pharmaceutical industry's digital transformation.

Full solution for this scenario: Pharma / drug industry solutions

This article was generated by AI. Customer cases are simulated scenarios based on real product capabilities and figures are illustrative; see product pages for official benchmarks.