The topic was to spot and potentially measure the amount of floating waste on the river Tisza and its offsets. It was a two-year long study in which a complete system was designed and built and performed measurements 50-70 % of waste dumped into rivers are plastics. Ranging from micro-plastics (< 0.1 µm) to macro-plastics (>5 cm). The nature of the plastic pollution depends greatly on the source of the pollution. In the river Tisza and its offsets, the pollution is mainly coming from landfills located near the upstream. In the first phase, an experimental motion-detection camera system was developed to try out multiple configurations during the research. The open-source motion software has been implemented, running on Raspberry Pi 3 as data collectors. The system uploaded data into a data server running in the cloud (Azure). The camera system was operating for more than a year and collected over 440,000 pictures. At the end of this phase, the conclusion was that individual plastic objects are not recognisable, only bigger groups of them. On top of this, we have seen that the optical noise is very high, rendering many of the pictures unfit for analysis, but the results still served as a very good starting point for the collection of AI training data. During the second phase, software was experimented. YOLOv3 and Faster R-CNN have been applied, eventually settling for Faster R-CNN with a ResNet-50-FPN base network.