A Fast Baseline System for Large Scale Bird Identification

We present a description of our approach for the “Bird task Identification LifeCLEF 2015”. Our approach consists of a baseline system based on the classification of Mel-bands representations of bird singing using a random forest classification. This setting proved to be fast during testing, extraction of Melbands and classification was done in a couple of hours. Our best system reached a Mean Average Precision of 14.5% and Recall of 14.5%.