Improving manual tagging of camera trap videos for wildlife studies: a visual assistance tool applied to the endangered huillin Lontra provocax.

Abstract

In the last decades, the use of camera traps for wildlife studies has increased significantly due to advancements in technology leading to cost reduction and improved reliability. They facilitate the study of wild animals in their natural environment in a minimally invasive manner, eliminating the need for researchers to remain in the field for long periods of time and to stay alert for extended hours. While their use is not too complex, data analysis often represents a bottleneck that demands considerable effort. When cameras are used to obtain videos, the first step upon retrieving information from them is to sort the videos, name them in a standardized manner, identify the content of each file, and label them to organize the information for subsequent data analysis. This process is laborious and repetitive, often leading to human errors that later hinder analysis. In this work an easy to use, customizable semi-automated video processing and tagging software is presented alongside camera trap data analysis from sites used by an endangered otter species Lontra provocax. With its use, we were able to process over 3000 video files in a short time, finding that in the area under study (Nahuel Huapi National Park) all otter sites were also used by American minks, foxes, domestic dogs, and cats. In a particular site a detailed analysis provided data showing a nocturnal pattern of use for the site. When compared to manual data processing the semi-automated process improved efficiency and overall user experience, and reduced the cognitive load associated with manual tagging. The aim of this work was to develop a versatile and user-friendly yet powerful solution to the typical tasks associated with the analysis of camera trap videos.