New Coake® Version 7.4.3 Released
This release introduces several major features and enhancements to the use of Coake. First and foremost, the new trevista® Smart Surface Inspection, based on Deep Learning, makes it possible to find and highlight defects in trevista images with little effort. Here, AI technology optimally complements classic image processing. It enables surface inspection solutions that could classically only be implemented with a great deal of effort or not at all. In order to annotate images, train neural networks and manage training models, appropriate tools are provided in Coake. Many of the existing commands (including filters and object recognition) have been extended to include a masking function with image or object masks. This makes it very easy to restrict the range of action of these commands to defined free-form areas. New commands have also been provided to improve the workflow when creating masks.
For the initial position tracking of the camera image, there is now also a general, powerful detection of the image alignment in addition to the commands of the Kronos package. Commands for processing timestamps, polygons, circular rings and image histograms are also new.
The default sorting of commands in the command bar has been significantly improved. Commands are now divided into logical steps and can therefore be found more easily.
For SDK users, the extension concept of Coake has been improved. A Coake project extension, mostly consisting of libraries, icons and translations, is now encapsulated in a single plugin folder.
The following are the major changes in detail:
1. trevista® Smart Surface Inspection
trevista® Smart Surface Inspection extends the broad capabilities of the trevista sensor family with another groundbreaking component - the defect image. This new dimension in surface inspection is based on current Deep Learning technology and simplifies the process of finding defects in trevista images. The new features can simplify existing solutions and also enable solutions where classical rule-based image processing reached its limits. trevista® Smart Surface Inspection is executable on both CPUs and GPUs. We recommend the use of the most powerful Nvidia graphics card possible, as well as the installation of the optional extension "Smart Surface Inspection GPU". trevista® Smart Surface Inspection is a paid add-on for Coake 7. A valid license enables you to save and apply the learned models. Managing and annotating image catalogs and training models (without saving) only requires a matching Coake license (standard or offline). The 31-day demo version of Coake still contains all possible functions, thus also a test license for trevista® Smart Surface Inspection.
2. Neural Evaluation
Among other things, Deep Learning technology is perfectly suited for the classification of data. In quality control, parts are classically classified into an IO and NIO class based on image data and rule-based checks. The new extension package "Neural Evaluation" based on current Deep Learning technology enables the evaluation of image data in IO and NIO only by pointing and training. This allows inspection tasks to be solved quickly and without in-depth image processing knowledge. Neural evaluation is executable on CPUs as well as on GPUs. We recommend using the most powerful Nvidia graphics card possible, as well as installing the optional "Addon Deep Learning" extension. The Deep Learning add-on is a paid add-on for Coake 7. A valid license enables you to save and apply the learned models. Managing image catalogs and training models (without saving) only requires a matching Coake license (standard or offline). The 30-day demo version of Coake still contains all possible functions, thus also a test license for Neural Evaluation.
3. Masks
In order to execute an image operation only on a partial area of the image, the image region/window was available up to now. The image region is defined in image coordinates and is always aligned with the pixel grid. To allow image operations within arbitrary shapes, this version introduces image and object masks. An image mask is a gray value image which contains binary information about active pixels. An object mask uses a geometric object (e.g. rectangle, circle, circular ring, polygon or segmented object) or a list of geometric objects to specify which pixels the operation should be applied to. Geometric objects are defined in the usual way in world coordinates, which is why they can also be used for layer tracking.
4. Command order in logical groups
The commands are now arranged in categories and groups in the command overview, which are modeled on the work steps when solving a BV project. This should result in a significant improvement of the workflow after a short period of getting used to it. Care has also been taken to ensure that all commands are directly accessible on a current standard monitor (16:9 Full HD).
5. SDK customizations and Coake extensions
The SDK is now installed by default next to the application under "C:\Programs\SAC\Coake 7.4.3 SDK". Before using the SDK, it should be copied to another location. This way the installed SDK always remains as a "clean state". Also there were some adjustments regarding the structure of Coake extensions. The individual components of extensions are now no longer distributed to individual folders in the project, but are encapsulated in a summarized plugin folder.
