Bird detection and monitoring in videos

Abstract

Conservation planning and conflict resolution require careful monitoring and analysis of wildlife. Video traps, in combination with AI-based analytic tools, are an excellent example of noninvasive technology used successfully in creating, planning, and assessing conservation policies. This research project aims to take a step further in understanding the behavior of birds by analyzing longer-term footage of large groups of birds, rather than using camera traps to collect still photos or short videos.The project’s primary goal is to create datasets, benchmarking tools, and unique algorithms that will aid in building automated video analysis tools to understand the individual and social behavior of birds.The proposed dataset poses several challenges in segmentation, detection, localization, and density estimation for cutting-edge vision techniques. Each shot contains an average of 207 birds, with a maximum count of 1500, making it a unique and complex dataset for computer vision algorithms. During the course of this open research project, the research team will scale this dataset for video segmentation and tracking, as well as create unique video analytics approaches for wildlife monitoring.The research project aims to provide a comprehensive understanding of bird behavior through video analysis. The footage will be analyzed to gain insights into the individual and social behavior of the birds. The large group of birds in each shot presents a unique challenge for computer vision algorithms, as traditional techniques struggle with segmentation, detection, localization, and density estimation. As a result, the we will develop new algorithms to overcome these challenges and create innovative video analytics approaches for wildlife monitoring.As part of this open science project, the dataset will be scaled for video segmentation and tracking, and new video analytics techniques will be developed for wildlife monitoring. The resulting dataset and benchmarking tools will serve as a valuable resource for future research on bird behavior and wildlife monitoring. This open science project will promote collaboration and knowledge sharing within the scientific community, leading to more effective conservation policies and conflict resolution strategies.