Deep Learning Case Study on Imbalanced Training Data for Automatic Bird Identification

Collisions between birds and wind turbines can be significant problem in wind farms. Practical deterrent methods are required to prevent these collisions. However, it is improbable that a single deterrent method would work for all bird species in a given area. An automatic bird identification system is needed in order to develop bird species level deterrent methods. This system is the first and necessary part of the entirety that is eventually able to, monitor bird movements, identify bird species, and launch deterrent measures. The system consists of a radar system for detection of the birds, a digital single-lens reflex camera with telephoto lens for capturing images, a motorized video head for steering the camera, and convolutional neural networks trained on the images with a deep learning algorithm for image classification. We utilized imbalanced data because the distribution of the captured images is naturally imbalanced. We applied distribution of the training data set to estimate the actual distribution of the bird species in the test area. Species identification is based on the image classifier that is a hybrid of hierarchical and cascade models. The main idea is to train classifiers on bird species groups, in which the species resembles more each other than any other species outside the group in terms of morphology (coloration and shape). The results of this study show that the developed image classifier model has sufficient performance to identify bird species in a test area. The proposed system produced very good results, when the hybrid hierarchical model was applied to the imbalanced data sets.