Recognition of species of Triglidae family using Deep Learning
Yakup Kutlu, Gokhan Altan, Bilal İşçimen, Servet A. Doğdu, Cemal Turan
Department of Computer Engineering, Facultyof Electric and Electronics, Iskenderun Technical University, Hatay, TURKEY
Institute of Natural and Applied Sciences, Mustafa Kemal University, Hatay, TURKEY
Kirikhan Vocational School, Mustafa Kemal University, Hatay, TURKEY
Faculty of Marine Sciences and Technology, Iskenderun Technical University, Hatay, TURKEY
This study is performed to classify fish species based on morphometric measurements between main points (fins, head and mouth) on fish image. Three species of Triglidae Family (Aspitrigla cuculus, Chelidonichthys lastoviza and Chelidonichthys lucernus) which has very similar in appearance of shape, color and the fin type are used for the classification. In the first stage, dataset was collected using images of fish species, and then morphometric features were extracted from the fish images using 13 landmarks. Deep Belief Networks is used as a classifier by combining with 3-fold cross validation method. Consequently, 3 fish-species of Triglidae family are separated with a high accuracy rate of 97.61%.
Keywords: Triglidae, species identification, morphometrics, Deep Learning, Deep Belief Networks, fish recognition