Deep Change Monitoring: A Hyperbolic Representative Learning Framework and a Dataset for Long-term Fine-grained Tree Change Detection

TL;DR

UAVTC is introduced, a large-scale, long-term, high-resolution dataset collected using UAVs equipped with cameras, specifically designed to detect individual Tree Changes (TCs).

Abstract

In environmental protection, tree monitoring plays an essential role in maintaining and improving ecosystem health. However, precise monitoring is challenging because existing datasets fail to capture continuous fine-grained changes in trees due to low-resolution images and high acquisition costs. In this paper, we introduce UAVTC, a large-scale, long-term, high-resolution dataset collected using UAVs equipped with cameras, specifically designed to detect individual Tree Changes (TCs). UAVTC includes rich annotations and statistics based on biological knowledge, offering a fine-grained view for tree monitoring. To address environmental influences and effectively model the hierarchical diversity of physiological TCs, we propose a novel Hyperbolic Siamese Network (HSN) for TC detection, enabling compact and hierarchical representations of dynamic tree changes. Extensive experiments show that HSN can effectively capture complex hierarchical changes and provide a robust solution for fine-grained TC detection. In addition, HSN generalizes well to cross-domain face anti-spoofing task, highlighting its broader significance in AI. We believe our work, combining ecological insights and interdisciplinary expertise, will benefit the community by offering a new benchmark and innovative AI technologies.

Figures (6)

• Figure 1: The proposed UAVTC is a dataset for long-term, precise tree monitoring aimed at supporting environmental protection. (a) UAVTC includes a large volume of high-resolution tree images captured by UAVs with cameras over one year. Tree changes encompass intrinsic (b) and extrinsic (c) patterns, which share hierarchical relationships. Traditional linear operators in Euclidean space are ineffective at representing these complex relationships. In contrast, hyperbolic spaces are naturally suited to embedding hierarchies with low distortion sarkar2011lowsala2018representation.
• Figure 2: The illustration of the data collection process. (a) the UAV equipped with a camera (DJI Zenmuse P1), (b) the flight routine outlining the test site, (c) the reconstructed test site after the UAV has captured the necessary imagery, and (d) the final cropped tree images.
• Figure 3: Framework of Hyperbolic Siamese Network. First, features extracted from the backbone Siamese network undergo a comparison process to compute change. Subsequently, a fully connected (FC) layer reduces the dimension of the change feature. The model utilizes exponential mapping to transform embeddings from the Euclidean space to hyperbolic space. A Hyp-BLR is followed for the classification Finally, a Hyp-BCE loss function is employed to train modes for TCD.
• Figure 4: The impacts of the embedding dimensions used in HSNs.
• Figure 5: The visualizations on Euclidean and hyperbolic spaces with t-sne van2008visualizing, respectively. ESN and HSN represent Siamese networks in Euclidean space and hyperbolic space, respectively.