Sick Deep Learning SensorApp Intelligent Inspection Plus released

Based on the InspectorP series of Sick 2D smart cameras, the deep learning SensorApp Intelligent Inspection Plus was officially launched, adding two deep learning functions of "Anomaly detection" and "Classification tools" on the basis of the traditional Quality Inspection Plus.

1. Why choose Sick Deep Learning

a) The Deep Learning-based Intelligent Inspection Plus SensorApp includes powerful "Anomaly Detection" and "Classification Tools", which enable applications that are not possible with traditional visual inspection.

b) We can provide customers with detailed application examples, support customers to train directly in the camera equipment or via the dStudio online platform, and the simple and user-friendly interface paves the way for rapid development for customers.

c) "Anomaly detection" and "Classification tools" can ensure that the analyte meets the quality inspection and classification needs, further improving efficiency, reducing costs and increasing customer satisfaction. In addition, the Quality Inspection Plus SensorApp, a traditional visual inspection tool, is also available in Intelligent Inspection Plus.

d) The SensorApp can be run in all InspectorP6xx 2D smart cameras in SICK, providing customers with a cost-effective solution that includes deep learning capabilities – no hardware needs to be used in addition to the InspectorP6xx 2D smart camera.

2.Anomaly detection

Anomaly detection tools can be used in complex applications where defective samples cannot be effectively predicted, such as surface inspection, weld inspection, gluing inspection, injection molding abrasive inspection, etc.

For training, you only need to collect good image samples. The anomaly detection tool will directly output the OK or NOK detection conclusion based on the training results of good samples, and display the defective areas in the form of a hot area map in the inspection image.

The anomaly detection tool enables customers to build simple and fast on-device applications, allowing users to train up to 100 good image samples in the InspectorP smart camera hardware.

Sick Deep Learning SensorApp Intelligent Inspection Plus released

　　In the anomaly detection function, the entire solution mainly includes three parts:

a) Use a smart camera to capture images in the field

b) Use a smart camera to train on the images

c) Real-time detection in the field using smart cameras

merit

a. There is no need to use an additional industrial computer (PC) for training, and fast training can be achieved inside the smart camera

b. Only a small number of positive samples need to be collected, i.e., the samples that are considered to be qualified only need a minimum of 2 samples and a maximum of 100 samples

c. There is no need to collect defective samples

d. There is no need to label the image

3.Classification tools

Classification detection tools can distinguish between visually similar objects and are suitable for complex applications such as changeable, unstable, reflective materials. It can also be easily implemented for assembly verification, defect classification, etc.

The tool works by collecting images and performing inspections in the InspectorP smart camera. In order to optimize accuracy and execution speed, the annotation, training, and evaluation processes are carried out using the SICK dStudio online platform. In this way, the number of samples collected can be larger.

　　In the classification function, the whole solution mainly includes three parts:

a) Use a smart camera to capture images in the field

b) The images were trained using the SICK deep learning cloud dStudio and the model was exported

c) Import the trained model to the smart camera and detect it in real time on site

　　4. dStudio

dStudio is SICK's online service platform for training classified neural networks, which is based on a variety of SICK smart camera hardware and further optimizes the training accuracy and execution speed.

dStudio has an easy-to-use interface that can be used by people without any prior knowledge of AI. Data such as training progress and training success rate are clearly displayed in the graph, and after the training is completed, the neural network needs to be downloaded and transferred to the smart camera hardware for production use. Click on the blue word for more information

　　5. Intelligent Inspection Plus

Product selection and scheme recommendation