Fish Image Classification Using Adaptive Learning Rate In Transfer Learning Method

The existence of fish species diversity in coastal ecosystems which include mangrove forests, seagrass beds and coral reefs is one the benchmarks determining health ecosystems. It certain that we must maintain, preserve care for so conservation efforts need to be carried out water areas. Many experts at Indonesian Fisheries Marine Research Development Agency often classify images manually, course it will take a long time, therefore with today's developments they can use latest technology. One reliable techniques terms image classification Convolutional Neural Network (CNN). As time goes by, course, many people want fast learning solving new problems faster better, transfer appears, adopts part CNN, name modified convolution layer. Observing needs field marine conservation, researchers decided solve this problem by using modifications. used an architectural model from pre-trained Mobilenet V2, known its light computing process applied our gadgets other embedded tools. research data 49.281 various sizes there are 18 types fish, pre-processing resize size 224x224 pixels. testing obtained accuracy score 99.54%, quite classifying images.