Pill counter: Automating real-time pill detection and quality control

In pharmaceutical production, precision is paramount. A single miscounted pill, misidentified medication, or foreign object in packaging can disrupt the entire supply chain. Traditional quality inspection methods rely heavily on manual control, which is time-consuming, prone to fatigue, and difficult to scale.

Honeycomb Software set out to create an AI-driven inspection solution to demonstrate how automated visual systems can reduce human error, enhance traceability, and ensure consistent product quality in real-time.

Our team designed a complete computer vision system, covering both software and physical hardware components. The engineers not only developed the AI model and interface but also built the physical prototype, including the camera mount and surface module, to simulate real-world inspection conditions.

The system analyzes live video streams from the mounted camera to detect, classify, and count pills automatically, providing immediate and visual feedback through an intuitive dashboard.

• Real-time pill recognition and classification: Detects multiple pill types by shape (round, capsule, oval, etc.) and color.
• Automated counting and visual breakdown: Displays total count, type distribution, and real-time results via the live dashboard.

• Anomaly detection: Instantly identifies foreign objects.
• Integrated physical prototype: Custom-built structure ensures stability, optimal lighting, and realistic data capture for computer-vision testing.

• Single-model architecture: Optimized for high accuracy and minimal latency, ensuring smooth real-time detection without data storage.
• Scalable design: Modular setup easily adapts to manufacturing, healthcare, or logistics pipelines.

By combining hardware engineering and AI expertise, Honeycomb Software delivered a fully functional end-to-end solution, from model training and UI design to physical assembly and testing.

The system achieved high accuracy in detecting and classifying multiple pill types, instantly identifying anomalies and processing continuous video streams on the spot. This demonstrated how computer vision can transform quality control into a faster, safer, and more reliable process.

By automating manual inspection and integrating real-time AI inference, the prototype proved the efficiency of Honeycomb Software’s approach to AI engineering, model optimization, system integration, and the ability to deliver lightweight, production-ready AI architectures that help organizations boost accuracy, ensure compliance, and accelerate decision-making across various business environments.