The general task of image classification seems to be solved due to the development of modern convolutional neural networks (CNNs). However, the high intraclass variability and interclass similarity of plankton images still prevents the practical identification of morphologically similar organisms. This prevails especially for rare organisms. Every CNN requires a vast amount of manually validated training images which renders it inefficient to train study‐specific classifiers. In most follow‐up studies, the plankton community is different from before and this data set shift (DSS) reduces the correct classification rates. A common solution is to discard all uncertain images and hope that the remains still resemble the true field situation. The intention of this North Sea Video Plankton Recorder (VPR) study is to assess if a combination of a Capsule Neural Network (CapsNet) with probability filters can improve the classification success in applications with DSS. Second, to provide a guideline how to customize automated CNN and CapsNet deep learning image analysis methods according to specific research objectives. In community analyses, our approach achieved a discard of uncertain predictions of only 5%. CapsNet and CNN reach similar precision scores, but the CapsNet has lower recall scores despite similar discard ratios. This is due to a higher discard ratio in rare classes. The recall advantage of the CNN decreases with increasing DSS. We present an alternative method to handle rare classes with a CNN achieving a mean recall of 96% by manually validating an average of 6.5% of the original images.

With recent advances in Machine Learning techniques based on Deep Neural Networks (DNNs), automated plankton image classification is becoming increasingly popular within the marine ecological sciences. Yet, while the most advanced methods can achieve human-level performance on the classification of everyday images, plankton image data possess properties that frequently require a final manual validation step. On the one hand, this is due to morphological properties manifesting in high intra-class and low inter-class variability, and, on the other hand is due to spatial-temporal changes in the composition and structure of the plankton community. Composition changes enforce a frequent updating of the classifier model via training with new user-generated training datasets. Here, we present a Dynamic Optimization Cycle (DOC), a processing pipeline that systematizes and streamlines the model adaptation process via an automatic updating of the training dataset based on manual-validation results. We find that frequent adaptation using the DOC pipeline yields strong maintenance of performance with respect to precision, recall and prediction of community composition, compared to more limited adaptation schemes. The DOC is therefore particularly useful when analyzing plankton at novel locations or time periods, where community differences are likely to occur. In order to enable an easy implementation of the DOC pipeline, we provide an end-to-end application with graphical user interface, as well as an initial dataset of training images. The DOC pipeline thus allows for high-throughput plankton classification and quick and systematized model adaptation, thus providing the means for highly-accelerated plankton analysis.