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

被引:10
作者
Eichinski, Philip [1 ]
Alexander, Callan [2 ]
Roe, Paul [1 ]
Parsons, Stuart [2 ]
Fuller, Susan [2 ]
机构
[1] Queensland Univ Technol, Fac Sci, Sch Comp Sci, Brisbane, Qld, Australia
[2] Queensland Univ Technol, Sch Biol & Environm Sci, Fac Sci, Brisbane, Qld, Australia
来源
FRONTIERS IN ECOLOGY AND EVOLUTION | 2022年 / 10卷
关键词
bird monitoring; ecoacoustics; deep learning; biodiversity; species recognition; active learning; CLASSIFICATION;
D O I
10.3389/fevo.2022.810330
中图分类号
Q14 [生态学(生物生态学)];
学科分类号
071012 ; 0713 ;
摘要
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.
引用
收藏
页数:11
相关论文
共 28 条
  • [1] Acevedo MA, 2006, WILDLIFE SOC B, V34, P211, DOI 10.2193/0091-7648(2006)34[211:UADRSA]2.0.CO
  • [2] 2
  • [3] Avian population consequences of climate change are most severe for long-distance migrants in seasonal habitats
    Both, Christiaan
    Van Turnhout, Chris A. M.
    Bijlsma, Rob G.
    Siepel, Henk
    Van Strien, Arco J.
    Foppen, Ruud P. B.
    [J]. PROCEEDINGS OF THE ROYAL SOCIETY B-BIOLOGICAL SCIENCES, 2010, 277 (1685) : 1259 - 1266
  • [4] COHN D, 1994, MACH LEARN, V15, P201, DOI 10.1007/BF00993277
  • [5] Preventing dataset shift from breaking machine-learning biomarkers
    Dockes, Jerome
    Varoquaux, Gael
    Poline, Jean-Baptiste
    [J]. GIGASCIENCE, 2021, 10 (09):
  • [6] The Good, the Bad, and the Risky: Can Birds Be Incorporated as Biological Control Agents into Integrated Pest Management Programs?
    Garcia, Karina
    Olimpi, Elissa M.
    Karp, Daniel S.
    Gonthier, David J.
    [J]. JOURNAL OF INTEGRATED PEST MANAGEMENT, 2020, 11 (01)
  • [7] Comparing recurrent convolutional neural networks for large scale bird species classification
    Gupta, Gaurav
    Kshirsagar, Meghana
    Zhong, Ming
    Gholami, Shahrzad
    Ferres, Juan Lavista
    [J]. SCIENTIFIC REPORTS, 2021, 11 (01)
  • [8] Deep Residual Learning for Image Recognition
    He, Kaiming
    Zhang, Xiangyu
    Ren, Shaoqing
    Sun, Jian
    [J]. 2016 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2016, : 770 - 778
  • [9] Fastai: A Layered API for Deep Learning
    Howard, Jeremy
    Gugger, Sylvain
    [J]. INFORMATION, 2020, 11 (02)
  • [10] Active learning for classifying long-duration audio recordings of the environment
    Kholghi, Mahnoosh
    Phillips, Yvonne
    Towsey, Michael
    Sitbon, Laurianne
    Roe, Paul
    [J]. METHODS IN ECOLOGY AND EVOLUTION, 2018, 9 (09): : 1948 - 1958