MOO: A Multi-view Oriented Observations Dataset for Viewpoint Analysis in Cattle Re-Identification

2026-03-04

AI Summary

This work addresses the challenge of cross-view (particularly aerial-to-ground) animal re-identification, which suffers from severe performance degradation due to drastic viewpoint variations. Existing datasets lack precise angular annotations, hindering systematic investigation of geometric factors. To bridge this gap, we introduce MOO, a large-scale synthetic cattle re-identification dataset comprising images of 1,000 distinct cows captured from 128 uniformly sampled viewpoints, featuring the first-ever precise viewpoint annotations. Leveraging MOO, we quantitatively analyze the impact of elevation angle on model generalization and identify a critical elevation threshold governing performance. Furthermore, we demonstrate that synthetic geometric priors derived from MOO effectively mitigate domain gaps. Experiments show that pretraining on MOO significantly enhances both zero-shot and supervised transfer performance across four real-world cattle re-identification datasets.

Technology Category

Computer Vision: Multi-modal VisionMachine Learning: Mixture of Experts (MoE)Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasetsUser Modeling, Personalization and Recommendation: Fairness-aware retrieval and rankingWeb Mining and Content Analysis: Large pretrained models with web data

Abstract

Animal re-identification (ReID) faces critical challenges due to viewpoint variations, particularly in Aerial-Ground (AG-ReID) settings where models must match individuals across drastic elevation changes. However, existing datasets lack the precise angular annotations required to systematically analyze these geometric variations. To address this, we introduce the Multi-view Oriented Observation (MOO) dataset, a large-scale synthetic AG-ReID dataset of $1,000$ cattle individuals captured from $128$ uniformly sampled viewpoints ($128,000$ annotated images). Using this controlled dataset, we quantify the influence of elevation and identify a critical elevation threshold, above which models generalize significantly better to unseen views. Finally, we validate the transferability to real-world applications in both zero-shot and supervised settings, demonstrating performance gains across four real-world cattle datasets and confirming that synthetic geometric priors effectively bridge the domain gap. Collectively, this dataset and analysis lay the foundation for future model development in cross-view animal ReID. MOO is publicly available at https://github.com/TurtleSmoke/MOO.

Problem

• animal re-identification
• viewpoint variation
• Aerial-Ground ReID
• geometric variation
• elevation change

Innovation

• synthetic dataset
• viewpoint analysis
• elevation threshold
• cross-view re-identification
• domain transfer