High Accuracy Identification of Bird Species using Convolutional Neural Network Algorithm

This research paper deals with the tactic to identify the bird species using Convolutional Neural Network (CNN) algorithm with different activation functions for classification of an image. There are differing types of scenarios present in birds like different sizes, angles, colours, and shapes from a human point of view. Compared to the supported the audio classification of birds, there is a robust variation present in images to identify the species of the bird. Also, it is very difficult for humans to identify the birds through images. The proposed system facilitates the differentiation and classification of birds by employing the principles of part detection and CNN feature extraction through the utilization of multiple convolutional layers. These features are then gathered and used to feed the classifier for categorization. A range of bird photos and activation functions are used to evaluate the CNN algorithm. The various testing data set like cifar-10, Caltech-UCSD Birds 200 are collected and therefore the nodes of the image compared with them to find the accuracy Python is used in Anaconda 3 or Spyder 3 software to build the suggested system. Convolutional Neural Network using the ReLU-Sigmoid activation function produces higher accuracy than CNN using other activation functions, according to the analysis and results of the experiment.