WildDepth: A Multimodal Dataset for 3D Wildlife Perception and Depth Estimation

Muhammad Aamir; Naoya Muramatsu; Sangyun Shin; Matthew Wijers; Jia-Xing Zhong; Xinyu Hou; Amir Patel; Andrew Loveridge

cs.CVMar 17, 2026

WildDepth: A Multimodal Dataset for 3D Wildlife Perception and Depth Estimation

Authors: Muhammad Aamir, Naoya Muramatsu, Sangyun Shin, Matthew Wijers, Jia-Xing Zhong, Xinyu Hou, Amir Patel, Andrew Loveridge,

Organizations: University of Oxford · University of Cape Town · University College London

Abstract

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. However, for the animal, in particular, the majority of 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.

Figures & tables

Figure 1: Reconstruction result for a walking Red hartebeest.

Datasets	Source	# Species	# Videos/Images	MultiModal
AnimalKingdom [ 21 ]	YouTube	850	30k	✗
MammalNet [ 6 ]	YouTube	173	20k	✗
KABR [ 11 ]	Drone	3	13k	✗
Meerkats [ 25 ]	Zoo	1	35	✗
ChimpACT [ 17 ]	Zoo	1	163	✗
MammAlps [ 9 ]	Camera Traps	5	6k	✗

Table 1: Publicly available datasets for wildlife monitoring

Panel of Figure 2 — Figure 2: Depth Maps by different MDEs for randomly selected frames from our Dataset

Model	RMSE $\downarrow$	AbsRel $\downarrow$	SILog $\downarrow$	$\boldsymbol{\delta<1.25}$ $\uparrow$
MonoDepth	2.52	0.213	0.317	0.76
DPT	2.42	0.192	0.301	0.80
DepthPro	2.18	0.176	0.278	0.83
DepthAnything	1.94	0.161	0.254	0.86

Table 2: Depth estimation performance comparison. Lower is better for RMSE, AbsRel, and SILog; higher is better for

\delta<1.25

Panel of Figure 4 — Figure 4: Behaviour recognition for randomly selected frames from our WildDepth dataset.

Figure 5: Examples from target point densification of walking Elephant (top) and Lion (bottom).

Figure 6: 3D reconstructions of walking Camel (top), Giraffe (middle), and Ankole cow (bottom).

Method	MAE ( $\downarrow$ )	$R^{2}$ ( $\uparrow$ )
Planar Depth	$0.13\pm 1.14$	$-0.10\pm 0.13$
Full (ours)	$0.16\pm 1.12$	$0.25\pm 1.36$

Table 3: Quantitative comparison for the Red Hartebeest sequence. While MAE is comparable, only our method achieves a positive

R^{2}

, indicating a meaningful geometric fit.

Figure 7: Distribution of per-frame evaluation metrics for the Red Hartebeest sequence. Only our Full method consistently achieves a positive

R^{2}

Depth Estimation

Depth estimation infers scene distance from images and sensors, enabling 3D reconstruction, navigation, and spatial understanding.

See all 198 papers

WildDepth-v2: A Synchronized RGB--LiDAR--Radar Dataset and Benchmark for Metric Depth, Motion, and Tracking of Wildlife
Oct 22, 2026Muhammad Aamir, Sangyun Shin, Naoya Muramatsu +4Metric Depth Estimation Wildlife Monitoring
Department of Computer Science, University of Oxford, Oxford, UK · University of Cape Town, Cape Town, South Africa · Wildlife Conservation Research Unit, Department of Biology, University of Oxford, UK +1
WildCross: A Cross-Modal Large Scale Benchmark for Place Recognition and Metric Depth Estimation in Natural Environments
Mar 2, 2026Joshua Knights, Joseph Reid, Kaushik Roy +3Visual Place Recognition Metric Depth Estimation
CSIRO Robotics, Data61, CSIRO, Australia. · Queensland University of Technology (QUT), Australia.
SAM 3D Animal: Promptable Animal 3D Reconstruction from Images in the Wild
May 8, 2026Xuyi Hu, Jin Lyu, Jiuming Liu +43D Reconstruction 3D Mesh Reconstruction
University of Cambridge · Southern University of Science and Technology · Tsinghua University +1

Oct 22, 2026J/K move · Enter open · S save

WildDepth-v2: A Synchronized RGB--LiDAR--Radar Dataset and Benchmark for Metric Depth, Motion, and Tracking of Wildlife

Muhammad Aamir, Sangyun Shin, Naoya Muramatsu, Matthew Wijers+3

Department of Computer Science, University of Oxford, Oxford, UK · University of Cape Town, Cape Town, South Africa · Wildlife Conservation Research Unit, Department of Biology, University of Oxford, UK+1

Metric Depth Estimation Wildlife Monitoring Radar-LiDAR Fusion Depth Completion Multi-Object Tracking

Open paper AskPDF arXiv

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.

Abstract

Depth Estimation

Explore similar work

WildDepth-v2: A Synchronized RGB--LiDAR--Radar Dataset and Benchmark for Metric Depth, Motion, and Tracking of Wildlife

WildCross: A Cross-Modal Large Scale Benchmark for Place Recognition and Metric Depth Estimation in Natural Environments

SAM 3D Animal: Promptable Animal 3D Reconstruction from Images in the Wild