Publication | Open Access
The use of mobilenet v1 for identifying various types of freshwater fish
27
Citations
0
References
2020
Year
Convolutional Neural NetworkEngineeringMachine LearningFreshwater FishVarious TypesImage ClassificationImage AnalysisData SciencePattern RecognitionAquacultureFishery ManagementEmbedded Machine LearningMachine VisionFeature LearningMachine Learning ModelObject DetectionComputer ScienceFish FarmingDeep LearningComputer VisionDeep Neural NetworksAquatic OrganismMobilenet V1
Abstract In recent years, the business opportunities for freshwater fish potential utilization are very promising. Freshwater fish including pomfret, Nile tilapia, carp, goldfish, tilapia fish, snapper fish, and catfish have great economic value. They are usually exported in living conditions to several countries such as Singapore, Japan, Hong Kong, Taiwan, and Malaysia. The business of fish particularly freshwater fish is one of the national income sources. As the forms of fish vary, it is important to distinguish the types of freshwater fish utilizing object detection. This detection can be done by implementing deep learning. MobileNet V1 is a deep learning model that can be used for object detection or image classification. MobileNet V1 can work on smartphones or other embedded devices that still produce high-level accuracy. In this study, MobileNet V1 was trained with learning rate parameters 0.0004 and epoch 20.000. The use of these parameters obtained an accuracy rate of 90% in the detection of types of freshwater fish.