SPECTRO by HACARUS – Smarter Visual Inspections

SPECTRO by HACARUS – Smarter Visual Inspections

While existing AVI/ AOI systems achieve high inspection performance, there is the problem of having many false positives. This results in a secondary inspection involving human labor. By using our visual inspection system, we can reduce the amount of human labor in the secondary inspection and automate it. SPECTRO is able to detect and classify defects in manufacturing on substrates, precision parts, metals, plastics and food among others. There are many AI models suitable for each work. Our visual system can be used as a single application or integrated in to an existing system as a module. By using the API and SDK, it makes easy to integrate the solution in to the inspection line currently in operation and to add AI function to your own product. When used as a single application, we provide a dashboard for administrators and an operation panel for workers as an interface. Enables quick inspection operations using tablets and patch panels

*Please note that member solutions are often customizable to meet the needs of individual enterprise end users.



  • Heat Map
    A Heat Map visualization highlights the extend to which parts of an object are contributing to the inspection result – providing a transparency and explainability of the results.
  • Auto-crop
    Automatically recognizes the inspection object and separates it from the background – leading to precise prediction results even when the object placement differs.
  • IoU- and Anomaly Score
    Indicates the likeliness of a defect and the precision of the located anomaly, allowing for efficient evaluation and tuning of inspection models.


WorldwideBrazilChina (PRC)FranceGermanyIndiaJapanKoreaMexicoOther - Asia PacificOther - Europe and AfricaOther - North and South AmericaRussiaTaiwanUnited KingdomUnited StatesAI Software/SaaSAutomotiveManufacturingAnomaly DetectionFactory AutomationCSP - Amazon Web ServicesIntel® Xeon® Scalable ProcessorIntel® Distribution for Python*Intel® Optimization for TensorFlow*Models can be trained - data input only requiredLinux ProprietaryOther