Editorial Type:
Article Category: Research Article
 | 
Online Publication Date: Nov 17, 2023

Evaluation of Deep Learning-Based Monitoring of Frog Reproductive Phenology

and
Page Range: 563 – 570
DOI: 10.1643/h2023018
Save
Download PDF

To evaluate the utility of a deep-learning approach for monitoring amphibian reproduction, we examined the classification accuracy of a trained model and tested correlations between calling intensity and frog abundance. Field recording and count surveys were conducted at two sites in Kyoto City, Japan. A convolutional neural network (CNN) model was trained to classify the calls of five anuran species. The model achieved 91–100% precision and 75–98% recall per species, with relatively lower performance on less abundant species. Computational experiments investigating the effects of the number and seasonality of the training samples showed that models trained on larger datasets from broader recording seasons performed better. Calling activity was high when males were abundant (Pearson’s r = 0.45–0.66), although correlations between the calling activity and the number of pairs in amplexus were generally weaker. Our results suggest that deep learning is an effective tool for reconstructing the reproductive phenology of male anurans from field recordings. However, caution is required when applying to rare species and when inferring female reproductive activity.

  • Download PDF
Copyright: © 2023 by the American Society of Ichthyologists and Herpetologists
Fig. 1.
Fig. 1.

Performance of the trained ResNet18 model. The values of (A) precision and (B) recall are shown for all species.


Fig. 2.
Fig. 2.

Effects of dataset size and seasonal extent on model (A) precision and (B) recall. Training samples are selected from either early one-third, early two-thirds, or all of the recording period. Lines connect average values.


Fig. 3.
Fig. 3.

Calling activity of the five anuran species in Kyoto City, Japan, inferred by the trained model. Darker blue represents a greater number of audio segments in which calls are detected, and gray color represents missing recordings.


Fig. 4.
Fig. 4.

Relationships between the number of males or pairs and calling activity, quantified by the daily average number of call segments. Black bars represent the numbers of males, and orange represent the numbers of pairs in amplexus. Pearson’s correlation coefficients (r) between abundance and calling activity are shown.


Contributor Notes

Department of Zoology, Graduate School of Science, Kyoto University, Sakyo, Kyoto 606-8502, Japan; Email: (KK) kaede.kmr@gmail.com; and (TS) sota.teiji.88u@st.kyoto-u.ac.jp. Send correspondence to KK.

Associate Editor: D. S. Siegel.

Received: Mar 07, 2023
Accepted: Sep 10, 2023