Publications

Abstract : Bird audio identification is one of the challenging fine-grained tasks due to various complexity in the signal. In the current work, we present a new bird audio dataset from the Indian subcontinent and propose a novel hybrid features-based ensembled residual convolutional neural network to identify bird audios from the Indian subcontinent. We utilized mel-frequency cepstral coefficients (MFCC) and melspectrogram features to train the neural network. We compared the results of the proposed model with other machine learning and deep learning models. The results show that our proposed model achieved the best accuracy of 92% and best F1-score of 91% on using modified ResNet50 model. The dataset and the experiemental codes are available at GitHub