A Convolutional Neural Network Bird Species Recognizer Built From Little Data by Iteratively Training, Detecting, and Labeling

Automatically detecting the calls of species of interest in audio recordings is a common but often challenging exercise in ecoacoustics. This challenge is increasingly being tackled with deep neural networks that generally require a rich set of training data. Often, the available training data might not be from the same geographical region as the study area and so may contain important differences. This mismatch in training and deployment datasets can impact the accuracy at deployment, mainly due to confusing sounds absent from the training data generating false positives, as well as some variation in call types. We have developed a multiclass convolutional neural network classifier for seven target bird species to track presence absence of these species over time in cotton growing regions. We started with no training data from cotton regions but we did have an unbalanced library of calls from other locations. Due to the relative scarcity of calls in recordings from cotton regions, manually scanning and labeling the recordings was prohibitively time consuming. In this paper we describe our process of overcoming this data mismatch to develop a recognizer that performs well on the cotton recordings for most classes. The recognizer was trained on recordings from outside the cotton regions and then applied to unlabeled cotton recordings. Based on the resulting outputs a verification set was chosen to be manually tagged and incorporated in the training set. By iterating this process, we were gradually able to build the training set of cotton audio examples. Through this process, we were able to increase the average class F1 score (the harmonic mean of precision and recall) of the recognizer on target recordings from 0.45 in the first iteration to 0.74.

Prey removal in cotton crops next to woodland reveals periodic diurnal and nocturnal invertebrate predation gradients from the crop edge by birds and bats

Factors influencing the efficacy of insectivorous vertebrates in providing natural pest control services inside crops at increasing distances from the crop edge are poorly understood. We investigated the identity of vertebrate predators (birds and bats) and removal of sentinel prey (mealworms and beetles) from experimental feeding trays in cotton crops using prey removal trials, camera traps and observations. More prey was removed during the day than at night, but prey removal was variable at the crop edge and dependent on the month (reflecting crop growth and cover) and time of day. Overall, the predation of mealworms and beetles was 1-times and 13-times greater during the day than night, respectively, with predation on mealworms 3-5 times greater during the day than night at the crop edge compared to 95 m inside the crop. Camera traps identified many insectivorous birds and bats over crops near the feeding trays, but there was no evidence of bats or small passerines removing experimental prey. A predation gradient from the crop edge was evident, but only in some months. This corresponded to the foraging preferences of open-space generalist predators (magpies) in low crop cover versus the shrubby habitat preferred by small passerines, likely facilitating foraging away from the crop edge later in the season. Our results are in line with Optimal Foraging Theory and suggest that predators trade-off foraging behaviour with predation risk at different distances from the crop edge and levels of crop cover. Understanding the optimal farm configuration to support insectivorous bird and bat populations can assist farmers to make informed decisions regarding in-crop natural pest control and maximise the predation services provided by farm biodiversity.