CONTACTMANUFACTURING
Simplify neural network training W
Simplify neural network training with Cognex’ VisionPro Deep Learning graphical user interface and intuitive programming environment
ith VisionPro Deep Learning, Cognex offers a deep learning-based image
analysis software specifically designed for factory automation. For tasks in industrial image processing that
can be solved by formulating rules, traditional machine vision systems are usually the right choice. However, these systems often reach their limits when the objects to be inspected introduce variability and cannot be easily automated using rules- based programming. This problem occurs in the inspection of food, in the evaluation of soldered or welded seams, and in many other applications. For use cases like these, machine vision systems based on deep learning technologies have established themselves as an innovative solution in recent years. VisionPro Deep Learning from Cognex significantly reduces this entry barrier through a graphical user interface that simplifies the training of the neural network for users.
applications that are too complex for traditional rule-based image processing approaches. Even with varying backgrounds or surface textures, the Red Analyse tool finds the smallest defects. In the example, the tool detects seam problems in textiles.
Identification with Blue Locate The Blue Locate tool is ideal for tasks where parts with different appearances need to be detected or counted. With its robust design, Blue Locate successfully identifies features on confusing backgrounds, lowcontrast parts, and even parts that bend, change shape, or are poorly lit. Even with variations in perspective, orientation, brightness, gloss or colour, Blue Locate reliably locates parts learned from sample images. For these reasons, Blue Locate is suitable for use in automated assembly verification, among other tasks.
is a robust classifier that can distinguish different classes of objects, identify defect types, and classify good and bad parts. Trained with a set of labeled images, Green Classify identifies objects based on their common features such as colour, texture, material, packaging, and defect type, and divides them into classes. In doing so, the tool tolerates natural variations within the same class and effectively distinguishes acceptable variants from different classes. Green Classify handles even complex classification tasks very quickly and does not require complicated and time-consuming programming.
Intuitive graphical training VisionPro Deep Learning tools are trained with images, unlike traditional image processing methods, which are programmed with rules-based algorithms. VisionPro Deep Learning's intuitive graphical user interface provides a simple environment for controlling and developing applications and significantly reduces the effort required to collect images, train the neural network, and test it on different sets of images. The Blue Locate tool is suitable for automated assembly or completeness checks, among other tasks, due to its ability to accurately identify different parts. Users can choose from four image analysis tools in VisionPro Deep Learning. They are optimised for vision inspections in factory automation applications and therefore only require a small number of images for rapid training. Blue Locate, Red Analyse, Green Classify and Blue Read tools solve
36 SEPTEMBER 2021 | ELECTRONICS TODAY
Detect defection with Red Analyse When the smallest defects must be found, despite different backgrounds and surface textures of parts, the Red Analyse defect detection and segmentation tool is the right choice. By training examples of good and bad parts, it is able to tolerate normal deviations in terms of appearance, while accurately detecting defects, impurities and other flaws. Red Analyse can also be used to segment variable areas in an image. Examples include welds, glued or painted areas, and background features that are dynamically hidden from the image to facilitate other inspections. The Green Classify tool is a robust classifier
that distinguishes different object types, identifies error types, and inspects and classifies images.
Classification with Green Classify VisionPro's Deep Learning tool, Green Classify,
Optical Character Recognition with Blue Read Reading and recognising fonts and codes is one of the most common tasks of image processing systems. Deformed, skewed or poorly etched characters can pose a real challenge, which can be easily mastered by using the Blue Read tool. The tool uses a pre-trained deep- learning font library and deciphers difficult characters on this basis. The Blue Read tool reliably deciphers and recognises even deformed, skewed or poorly etched fonts and codes. The user-friendly GUI eliminates the need
for complex programming and drastically reduces development time: users only need to define the target area, set the character size, and mark the characters in the images. In just a few steps, the robust tool can be trained to read application-specific plain text that cannot be decoded by conventional OCR tools. Plus, training can be completed without any image processing or deep learning knowledge. In addition, the optical debug function detects misread characters, which can then be easily corrected. One major benefit of VisionPro Deep Learning is the ability to combine the available tools. This allows complex problems to be broken down into smaller individual steps to simplify project optimisation and reduce the number of training images required.
www.cognex.com
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58
