Bird Recognition in Urban Ecological Areas Based on Deep Learning

With the swift development of technology, image recognition based on deep learning has become known to a wider public and grows mature gradually, which can be used in different fields such as face, vehicle, animal and plant recognition. This research focuses on bird image classification and recognition in urban ecological areas which constructs ResNet-50 model to classify and identify 20 species of birds. It summarises the existing research relevant to deep learning and introduces theoretical principles of ResNet-50 model. It aims to improve the efficiency and accuracy of image recognition. Based on this research, ornithologists can make better and more specific strategies for bird protection according to different species in order to improve ecological environment. As for the results, the accuracy and loss ratios of every training session are calculated and displayed using line charts. Besides, all bird images are marked as the name of its species and are determined if the label is true. The research achievement can be applied to the development of other fields with the recognition techniques of images.