Recognition of Bird Species Using Multistage Training with Transmission Learning

Object localization is a computer vision technique to identify real-world objects such as birds, cats, flowers, cars in images or videos. The algorithm is based on a feature extraction and learning algorithm to recognize instances of an object category. Bird’s species are the most amazing creature exist on earth. They are sensitive to changes in the environment and hence acts as bioindicator species. The main aim of this project is to identify bird species from a high-resolution digital image of Himalayan birds which would help beginner bird watchers or general people for identification. The data sets for the identification of birds are provided by Kaggle which consists of 16 species of birds. For the reduction of the overfitting problem, a data augmentation process is implemented. The model achieves an accuracy of 50.64 or 0.5064% on the dataset of Kaggle.