We are thrilled to present PEKAT VISON 3.14, the latest version of our cutting-edge deep-learning visual inspection software. In the last few months, we have been working on multiple improvements of PEKAT VISION that are part of the latest release, v. 3.14, setting a new standard for precision, efficiency, and reliability in automated inspection systems, available for download here.
In today’s fast-paced industrial landscape, ensuring the highest level of quality control is paramount. Manufacturing processes demand meticulous inspection to detect even the slightest defects, inaccuracies, or deviations that can compromise product integrity. This is where PEKAT VISION 3.14 shines, offering an unparalleled solution to streamline and enhance visual inspection across a wide range of industries.
True to our promise of providing an intuitive and user-friendly software environment with no or limited need for programing, depending on the use case, we increased the number of functions requiring no programming. The software user interface empowers operators of all technical background to effortlessly train and manage the inspection parameters, making it an accessible tool for both experienced professionals and newcomers to visual inspection technologies.
We focused on speed and accuracy and we have also listened to your feedback. Among the many improvements introduced in version 3.14 we would like to highlight the following:
Smarter Merging of Branches in Parallelism after Preprocessing
Cut, Scale, Resize, etc. now merge into the original branch in parallelism without the need for any coding, making preprocessing in a parallel branch much easier and accessible.
Significant Increase of Training and Inference Speed of the OCR and Anomaly modules
The speed of both, the training and inference has significantly increased. We now offer two modes of anomaly detection: Standard (formerly Precise) and Lightweight (formerly Fast).
Multiple Tags for Images
Multiple tags (such as “training”, “good”, “bad”, for example) can now be assigned to an image, enabling easier filtration and annotation of images
Auto select newly trained model – automatically selects the model after training and switches to the model detail
Updated visualization of annotations – improved rectangle design, easier manipulation and new keyboard shortcuts
Select multiple models for deletion – multiple models can be selected for deletion enabling faster deletion of models in a project with many models
Loss number in model detail – loss number is displayed under the loss curve during and after training
Added model number to model name – a training number is automatically added to a model during training, making comparison of training models easier
Complete processing time in debugger – total image processing time of all models is now displayed in debugger