Automatic fish sorting using image processing is an important part of making the fishing industry more productive through automation. In a production context, fish sorting is generally done manually by workers, which consumes time and energy, especially because fish sorting still involves manual actions. Therefore, researchers designed and built a fish sorting tool based on type and quality in the fishery product management process, designed a fish quality type system that could work based on eye color, scale or skin color parameters as aspects of the YOLO algorithm, and tested the accuracy level of the sorting tool's success. fish in detecting the type and quality of fish based on image processing. This research applies the Research and Development method for research continuity. In the research results obtained from the system this time to identify the type from 16 fish samples, the reading success rate was 100%. Meanwhile, the results of identifying fish quality from 8 images of test data from two types of fish obtained a success rate of 100%. And the success rate obtained in the actuator test was 100%. For the AI training results, the precision (non-background) value was 0.98 (98%), recall (non- background) 0.96 (96%) and F1 score (non-background) was 0.97 (97%). It is hoped that this research can be a guide for automatic fish detection, classification and sorting of fish.
Keyword: Automatic sorting tools, Image processing, YOLO, Fish quality, Fish
types.