Adding depth to camera trap analyses

Measuring animal distance and motion in metric units underpins abundance estimation, morphometrics, and movement ecology, yet monocular video lacks metric scale and wildlife datasets lack metric ground truth. We present WildDepth 2, extending the synchronized RGB-LiDAR WildDepth dataset in three directions. (i) Imaging radar: 37 RGB-LiDAR-4D-radar sequences (19,716 frames) with per-point Doppler velocity and a measured error budget, to our knowledge the first imaging-radar wildlife recordings. (ii) Benchmarking: a frozen, checksummed hidden test set for RGB-only distance estimation and RGB+sparse-LiDAR depth completion, evaluated separately on animal and background pixels. (iii) Motion and tracking: metric 3D trajectories obtained by lifting 2D tracks using LiDAR, radar, or monocular depth. Across 48,723 animal frames, object-level radar calibration improves monocular depth over scene-level LiDAR calibration without requiring LiDAR at inference (AbsRel 0.120 vs. 0.198), although animal-level LiDAR remains more accurate (0.052). Across 431 tracks, once-per-track radar scaling reduces 3D position error four-fold (2.20 to 0.58 m) and distance-sampling detection-function bias from −13% to −2%, matching once-per-track LiDAR calibration. Transferring radar Doppler onto LiDAR produces motion masks 3.7 times denser than radar alone at 72% precision, reduces moving-animal LiDAR accumulation error by 37%, and estimates radial speed (r = 0.60), albeit with approximately 50% underestimation.

@article{aamir2026wilddepthv2,
  title   = {WildDepth-v2: A Synchronized RGB--LiDAR--Radar Dataset and Benchmark for Metric Depth, Motion, and Tracking of Wildlife},
  author  = {Aamir, Muhammad and Shin, Sangyun and Muramatsu, Naoya and Wijers, Matthew and Patel, Amir and Loveridge, Andrew and Markham, Andrew},
  journal = {arXiv preprint arXiv:2610.31123},
  year    = {2026},
  doi     = {10.48550/arXiv.2610.31123}
}

Depth estimation and 3D reconstruction have been extensively studied as core topics in computer vision. Starting from rigid objects with relatively simple geometric shapes, such as vehicles, the research has expanded to address general objects, including challenging deformable objects, such as humans and animals. In particular, for the animal, however, the majority of the existing models are trained based on datasets without metric scale, which can help validate image-only models. To address this limitation, we present WildDepth, a multimodal dataset and benchmark suite for depth estimation, behavior detection, and 3D reconstruction from diverse categories of animals ranging from domestic to wild environments with synchronized RGB and LiDAR. Experimental results show that the use of multi-modal data improves depth reliability by up to 10% RMSE, while RGB–LiDAR fusion enhances 3D reconstruction fidelity by 12% in Chamfer distance. By releasing WildDepth and its benchmarks, we aim to foster robust multimodal perception systems that generalize across domains.

@inproceedings{aamir2026wilddepth,
  title        = {WildDepth: A Multimodal Dataset for 3D Wildlife Perception and Depth Estimation},
  author       = {Aamir, Muhammad and Muramatsu, Naoya and Shin, Sangyun and Wijers, Matthew and Zhong, Jia-Xing and Hou, Xinyu and Patel, Amir and Loveridge, Andrew and Markham, Andrew},
  booktitle    = {Computer Vision -- ECCV 2026},
  series       = {Lecture Notes in Computer Science},
  volume       = {17042},
  pages        = {40--58},
  year         = {2026},
  publisher    = {Springer},
  doi          = {10.1007/978-3-032-37422-6_3}
}

Gait is a distinctive behavioral characteristic that enables non-invasive individual identification without requiring physical interaction with an animal. While gait-based analysis has been extensively studied in humans, its application to wildlife remains limited due to environmental variability and the lack of scalable identification methods. This paper presents a fully automated, video-based pipeline for wildlife gait analysis and individual identification using deep spatiotemporal representation learning. The proposed pipeline uses the Segment Anything Model 3 (SAM3) to generate high-quality RGB and binary silhouette masks, robustly isolating animals from complex natural backgrounds. Segmented video sequences are processed using a convolutional neural network (ResNet18) for spatial feature extraction and a transformer-based video model (VideoPrism) for temporal motion modeling. Both models are fine-tuned using a classification objective and subsequently used as feature extractors to generate discriminative gait representations. Cosine similarity is then used to compare gait signatures, enabling similarity-based clustering of individuals without reliance on physical markings or invasive tagging. Experiments conducted on multi-source wildlife video data across multiple species demonstrate strong intra-individual consistency and clear inter-individual separation. Quantitative results using cosine similarity distributions and silhouette scores confirm the effectiveness of the proposed method. These findings demonstrate that gait dynamics provide a viable, non-invasive approach for individual identification in wildlife and highlight the potential of video-based deep learning pipelines for scalable ecological monitoring.

@article{aamir2026gait,
  title   = {A non-invasive video-based method for individual identification of wildlife using gait dynamics},
  author  = {Aamir, Muhammad and Wijers, Matthew and Shin, Sangyun and Loveridge, Andrew and Markham, Andrew},
  journal = {arXiv preprint arXiv:2607.04518},
  year    = {2026},
  doi     = {10.48550/arXiv.2607.04518}
}

Obtaining accurate metric measurements from 2D images captured by camera traps is valuable for ecological research, particularly population density estimation and morphometric analyses. Traditional methods require extensive manual calibration and image processing, making large-scale studies impractical. In this research study, we present a robust and automated pipeline that integrates monocular depth estimation (MDE) models with automated calibration and animal detection methods. Our pipeline uses deep learning based MDEs, including Dense Prediction Transformers (DPT), MiDaS 3.1, DepthAnything, DepthPro and UniDepthV2, along with multiple calibration methods such as linear, polynomial, and piecewise regression, to improve distance estimation accuracy. We introduce an automated mask generation and cross-camera calibration process to reduce manual effort. The proposed pipeline was validated using controlled ground truth data collected in Oxfordshire, UK, and subsequently tested on real-world wildlife data from the Bubye Valley Conservancy, Zimbabwe. Our proposed technique using UniDepthV2 MDE and piecewise linear regression for calibration reduces overall distance estimation error by up to 35% compared to existing approaches. Furthermore, we implemented height estimation as an indirect validation mechanism, presenting a correlation between true and estimated heights for different animals. Our fully automated pipeline advances camera trap-based wildlife monitoring, reducing human effort while improving scalability and reliability for conservation research.

@article{aamir2025robust,
  title   = {A robust metric distance and height estimation pipeline for wildlife camera trap imagery},
  author  = {Aamir, Muhammad and Wijers, Matthew and Loveridge, Andrew and Markham, Andrew},
  journal = {Ecological Informatics},
  volume  = {92},
  pages   = {103520},
  year    = {2025},
  doi     = {10.1016/j.ecoinf.2025.103520}
}

Understanding and monitoring wildlife behavior is crucial in ecology and biomechanics, yet challenging because of the limitations of current methods. To address this issue, we introduce WildPose, a novel long-range motion capture system specifically tailored for free-ranging wildlife observation. This system combines an electronically controllable zoom-lens camera with a LiDAR to capture both 2D videos and 3D point cloud data, thereby allowing researchers to observe high-fidelity animal morphometrics, behavior and interactions in a completely remote manner. Field trials conducted in Kgalagadi Transfrontier Park (South Africa) have successfully demonstrated WildPose's ability to quantify morphological features of different species, accurately track the 3D movements of a springbok herd over time, and observe the respiratory patterns of a distant lion. By facilitating non-intrusive, long-range 3D data collection, WildPose marks a significant complementary technique in ecological and biomechanical studies, offering new possibilities for conservation efforts and animal welfare, and enriching the prospects for interdisciplinary research.

@article{muramatsu2025wildpose,
  title   = {WildPose: a long-range 3D wildlife motion capture system},
  author  = {Muramatsu, Naoya and Shin, Sangyun and Deng, Qianyi and Markham, Andrew and Patel, Amir},
  journal = {Journal of Experimental Biology},
  volume  = {228},
  number  = {5},
  pages   = {JEB249987},
  year    = {2025},
  doi     = {10.1242/jeb.249987}
}