Bird Detection using CNN on Audio Classification Datasets
摘要
Accurate in ecological studies, birds plays important role in environment it reflects in biodiversity, there are two types of methods audio and images to identify bird species but images classification method is difficult to identity birds so that decided to give priority to audio based representation, we use light weighted Convolutional Neural Network (CNN) framework design to tackle audio based challenges task, we are working with imbalanced large dataset of 10,310 audio files across of 44 bird species. Our model train on 10 epochs and get testing accuracy of 88% and macro-average F1-score of 87%.