Deep Learning Inspection App for Vision Sensors

Enabling 2D vision sensors for complex sorting and quality inspection tasks

  • by SICK AG
  • September 10, 2021
  • 1283 views
  • Deep Learning Inspection App for Vision Sensors
    Deep Learning Inspection App for Vision Sensors

The SICK Intelligent Inspection Deep Learning SensorApp is now available to run directly onboard all InspectorP6xx vision cameras. SICK Intelligent Inspection is available as a seamless extension to the pre-installed Quality Inspection SensorApp, on all InspectorP6xx cameras. By combining traditional machine vision for quality inspection with a powerful extended Deep Learning capability, Intelligent Inspection opens up opportunities for users to automate challenging inspections that have not been possible previously. Vision classifications using Artificial Intelligence are now simple to set up and run across the entire range of SICK InspectorP6xx vision sensors. The newly launched, ultra-compact InspectorP61x is the smallest vision sensor currently available with Deep Learning running directly onboard, and the Intelligent Inspection capability extends right up to the rugged InspectorP65x with its extra high resolution and extended field of view. 

Practical and Affordable AI Classification

Where it has previously been very challenging to achieve consistently robust and repeatable quality inspections, with the SICK Intelligent Inspection SensorApp they can now be mastered with high levels of reliability and availability. Automation is therefore practical and affordable for complex imaging tasks such as checking the orientation of timber profiles by recognizing their annual ring structure; inspecting highly reflective surfaces, such as assemblies containing metal parts; classifying objects with slight differences within one class, such as food produce; or inspecting the integrity of solder in surface mount assemblies. 

Intuitive Step-By-Step Process

With the Intelligent Inspection image collection tool, users begin by collecting example images of their product in real production conditions. Then, they simply upload the images to SICK’s cloud-based training service, dStudio and use a step-by-step process to train and evaluate a neural network to meet the needs of the inspection. Additional images can then be added and evaluated, if needed, to perfect the result. Once satisfied, users deploy the custom-trained deep learning neural network to the SICK InspectorP6xx camera where it can begin to take decisions automatically, with no further Cloud connection necessary. The image inference is carried out directly on the device, so there is no need for an additional PC. As the system training is done in the Cloud, there is also no need for separate training hardware or software, saving on implementation time and cost.

With the benefit of training a neural network based on real examples, users can try out the suitability of deep learning classification for their application before purchasing the additional license required. They can also use traditional rule-based vision tools together with deep learning to solve the application. Developers working in SICK's AppSpace can save coding time and effort by plugging in to the SICK Nova machine vision toolbox to customize or create their own SensorApps. 

The SICK InspectorP6xx series of versatile 2D configurable and programmable vision sensors are designed for ease of use and versatility, whatever the application. From the miniscule InspectorP61x and the compact P62x, versatile InspectorP6xx vision sensors offer a scale of performance levels and longer ranges to meet every application. Characterized by high-quality lenses and powerful onboard LED illumination, all InspectorP6xx sensors have the SICK Quality Inspection SensorApp pre-installed onboard, and are supported by the versatility of SICK’s AppSpace software platform. For novice and expert users alike, the SICK Inspector P6xx family is founded on accessibility and on the flexibility of scalable onboard software to solve wide-ranging 2D machine vision applications.