The Fine-Grained Recognition of Bird Images Based on Joint Semantic Components and ResNet

Yunnan Province is famous with the largest recorded bird species in China, which is covering 848 species, totaling 1074 species and subspecies and belonging to 19 orders and 69 families. The bird resources in Yunnan Province have great ornamental value and economic value, and they are characterized by abundant species and diverse forms of birds. These variable and colorful features bring many challenges to their intelligent recognition and efficient management. At present, bird recognition and management mainly rely on manual methods and experience accumulation, which is inefficiency. Based on the latest artificial intelligence technology and the YUB-300-2019 bird dataset which we have developed, the fine-grained recognition model of bird images based on joint semantic components and ResNet is used. The experiment results showed our method has higher accuracy and can be generalized the wider bird identification. The method directly serves the society and acts on the original bird data, thus avoiding the difficulties and problems of feature extraction manually.