Conference Agenda
Overview and details of the sessions of this conference. Please select a date or location to show only sessions at that day or location. Please select a single session for detailed view (with abstracts and downloads if available).
Please note that all times are shown in the time zone of the conference. The current conference time is: 24th Aug 2026, 05:32:50am America, Santiago
|
Daily Overview |
| Session | |
|
Work in Progress (WP) In-Person Location: Room 07: Antartica | |
    | |
Region-based convolutional neural networks for automatic detection of a Neotropical bat species Universidad de Los Llanos - (CO), Colombia Bats use ultrasonic echolocation signals to orient themselves and feed, and analyzing these signals can be used for non-invasive, acoustic monitoring. Traditionally, this process is carried out manually by experts, which is a slow process due to the large volume of data and the temporal resolution of the signals. Currently, acoustic monitoring is an essential tool in ecology, especially for the study of bats, due to its non-invasive nature and its broad spatial and temporal coverage. For this reason, artificial intelligence techniques, including deep learning, have been incorporated into this work. This paper presents the use of region-based convolutional neural networks using a Faster R-CNN model for the automatic detection of pulse sequences from bat species in ultrasound recordings from spectrograms. A preliminary model showed high completeness values (>89%) and solid performance at mAP@50 (>892%), demonstrating adequate spatial localization capacity for the sequences, even in high biodiversity environments. However, lower accuracy and confusion values were observed between acoustically similar species, probably due to simultaneous vocalizations and the limited diversity of the training set. The feasibility of this approach for automated acoustic monitoring is evident, and future work should focus on expanding the dataset and improving interspecies discrimination. | |