CVMay 10, 2021

Overcoming the Distance Estimation Bottleneck in Estimating Animal Abundance with Camera Traps

arXiv:2105.04244v227 citations
Originality Incremental advance
AI Analysis

This addresses a critical problem for conservationists and ecologists by automating a labor-intensive step in estimating animal populations, though it is incremental as it builds on existing depth estimation methods.

The study tackled the bottleneck of manual distance estimation in camera trap surveys for animal abundance by developing an automated pipeline using monocular depth estimation and calibration, reducing manual effort by over 21 times.

The biodiversity crisis is still accelerating, despite increasing efforts by the international community. Estimating animal abundance is of critical importance to assess, for example, the consequences of land-use change and invasive species on community composition, or the effectiveness of conservation interventions. Various approaches have been developed to estimate abundance of unmarked animal populations. Whereas these approaches differ in methodological details, they all require the estimation of the effective area surveyed in front of a camera trap. Until now camera-to-animal distance measurements are derived by laborious, manual and subjective estimation methods. To overcome this distance estimation bottleneck, this study proposes an automatized pipeline utilizing monocular depth estimation and depth image calibration methods. We are able to reduce the manual effort required by a factor greater than 21 and provide our system at https://timm.haucke.xyz/publications/distance-estimation-animal-abundance

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes