Tablet and Vial Visual Defect Detection

• Production Manager, Solid Orals or Injectables
• Head of Quality Assurance
• Visual Inspection Supervisor
• Plant Engineering Manager

Tablets are inspected after compression and coating, and injectable vials and ampoules are inspected for particles and container defects before packing. Much of this is still done by trained inspectors at booths or sorting belts, followed by AQL sampling, and fatigue makes results vary across a shift.

We build defect detection models trained on images of your own products. For tablets, the model classifies chips, cracks, capping, black spots, embossing faults and coating defects on a sorting belt or inspection drum. For vials and ampoules, it checks cap and crimp, fill level, glass cracks and visible particles from images captured on your inspection machine or a dedicated camera station. Each unit gets a defect class and a confidence score. Clear rejects are removed, uncertain units go to a manual inspection lane, and every reject image is stored with its class for trending.

• Classifies tablet defects such as chips, cracks, capping, spots, embossing faults and coating defects
• Inspects vials and ampoules for cap and crimp faults, fill level, glass cracks and visible particles
• Sends low-confidence units to a manual inspection lane instead of forcing a decision
• Builds a labelled defect library from your rejects for qualification and retraining
• Trends defect types by batch, machine, punch set or filling line
• Works with images from existing inspection machines where they can be exported

1. 01Build the defect libraryInspectors label images of good units and known defects from your products to create training and challenge sets.
2. 02Train and challenge testThe model is trained and tested against a fixed challenge set before it sees live production.
3. 03Capture and classifyCameras capture each tablet or container, and the model assigns a defect class and confidence score.
4. 04Route by confidenceClear passes continue, clear rejects are removed, and uncertain units go to the manual lane.
5. 05Inspector decisionA trained inspector checks uncertain units, and the decision is added to the labelled library.
6. 06Trend and reportDefect counts and images per batch go to production and QA for investigation and product quality review.

1. Image capture
   • Area or line-scan cameras with controlled lighting
   • Spin-and-stop presentation for vials
   • Encoder-based triggering
2. Inference at the line
   • Edge GPU
   • Defect classification model
   • Confidence-based routing
3. Model management
   • Labelled defect library
   • Training and challenge test pipeline
   • Model version registry
4. Review and quality data
   • Manual inspection lane interface
   • Reject image archive
   • Defect trend dashboard

Trained inspectors decide every uncertain unit and confirm the labels used for retraining. QA approves each new model version after challenge testing, and AQL sampling continues as your batch-level check.

• Inspection machines or sorting systems with image export
• Line PLC for reject actuation
• MES or ERP for batch and machine data
• QMS for deviations and CAPA
• Data platform for product quality review trending

• Challenge sets with known critical defects run at start-up and after any change
• Model updates only through change control with documented test results
• Uncertain units always go to manual inspection
• No retraining on labels an inspector has not reviewed
• Critical defect classes tuned to avoid misses, accepting more manual checks

• Supports pharmacopoeial expectations for visible particles in injections, such as USP chapters 790 and 1790 and Ph. Eur. 2.9.20, alongside your AQL programme.
• Models and inspection recipes are validated as part of a GMP computerised system, with versioning and audit trails aligned to EU GMP Annex 11 and 21 CFR Part 11.

• Pharma & Life Sciences

  Packaging and Label Inspection

  Camera checks on cartons, labels and blister foils that verify batch, expiry, MRP and codes, compare print against approved artwork, and reject mismatches with image evidence.

• Pharma & Life Sciences

  Batch Manufacturing Record Review

  Pre-review of executed batch manufacturing and packaging records that flags blank fields, out-of-limit values, signature gaps and unlinked deviations, with page references, before QA starts.

• Manufacturing

  Visual Quality Inspection on Production Lines

  Cameras and vision models at the inspection station flag scratches, dents, missing features and print errors, divert clear rejects and hold borderline parts for an inspector.

• Pharma & Life Sciences

  Validation and Security Testing of AI Tools in GxP Settings

  Risk-based testing of LLM and ML tools used in quality, manufacturing or pharmacovigilance: accuracy, prompt injection, data leakage and audit-trail checks, packaged as validation evidence.