Is there any specific reason that most of the examples have to do with cells and microscopic images? It seems to me that KNIME would be suitable for a much wider range of applications. For example computer vision tasks: evaluating object recognition, semantic segmentation, face detection, pedestrian tracking etc. Or are these algorithms too heavy-weight and people prefer to just code them in C++ without a workflow tool like KNIME? Another guess of mine would be that computer vision people experiment a lot with new algorithms and want to change the algorithm here and there all the time and the personal time-overhead of having to write nodes instead of easy-to-maintain succint classes is just too large. In contrast, processing medical cell images has more established methods. Maybe the attractive aspect of KNIME is that non-programmer people can create algorithms easily, but computer vision people can program themselves most of the time.
Or maybe it's just tradition or insufficient marketing in other application fields.
that's really a good question! But the reason for that is fairly simple: we (Christian and I, the responsible persons for the image processing extension) doing our phd at the chair of Bioinformatics and Information Mining at the University of Konstanz where also KNIME was initiated and originally developed. Hence, during our work we are mostly concerned with image processing tasks from Biology, Chemistry and sometimes Physics, whose images mostly stem from microscopes.
But that doesn't mean that KNIME and the image processing extension can't be used for other purposes. We really would like to see KNIME and the Image Processing Plugin being applied to other (non-microscopic) fields! But we from our side just have no time to provide other sophisticated uses cases and heavy-weighted algorithms (e.g. for face detection) beyond microscopy. But as soon as somebody wants to have their image processing algorithms being available in KNIME we will be very happy to help and make this happen.
Best,
Martin
P.S.: The tesseract integration (OCR) is at least one example that has nothing to do with microscopy ;)
one comment in addition: we have for example a non-public prototype of an OpenCV integration in KNIME (JavaCV) where we did some object recognition and so on. So it's certainly doable, but we didn't have usecases yet.