Bird image recognition and classification using Watson visual recognition services from IBMCloud and Conventional Neural Network (CNN)

Abstract

The birdwatchers and pepole admiring the beauty of birds search in the books and the encyclopedias to identify species and provide information that characterizes eash bird but this solusion is not pratical, we proposed to develope an android platform named Birds Predictor to assist users in recognizing about 30 species of endemic birds in the world. Bird images are injected in a convolutional neural network (CNN) to localize prominent features. First, we create an image generator for the training data. Then, we load training images. After that, we create a neural network and the convolutional layer. Finally, we load the unknown bird image and applied the argmax function to get a probability of bird features. To identify the images downloaded or captured by mobile users the results of the parameters learned from the characteristics of the birds are used. For the Mobile Application we use IBM Cloud that offers the possibility to store a lot of data and trains it using the visual recognition service, then we send the image that we want to predict its type from our android application. We just connect the IBM project that contains the training images with our Android Studio project using an API Key, and IBM process classifies the image captured or uploaded from the application and returned the type of bird.