BarkVisionAI: Novel dataset for rapid tree species identification

Tree species identification and mapping is crucial for forest management, biodiversity conservation, and ecological research. Bark images can be captured easily from the ground-level and can provide large amount of information about the tree species and its health. Yet, existing datasets for tree bark images are often limited in scope, lacking diversity in species representation and temporal attributes. To address these limitations, we present BarkVisionAI, a comprehensive dataset of 167361 tree bark images for 13 species collected from diverse forest types across India. Each image is labeled with species name, device attributes, and timestamp, providing a robust foundation for studying species identification and the variability of bark characteristics. We are providing detailed metadata information about each image, encouraging its use in ecological research, machine learning model training, and environmental monitoring. Benchmarking experiments using standard image classification models demonstrate the dataset’s utility and effectiveness, highlighting its potential as a valuable resource for developing robust, real-world applications in automated tree species identification and environmental change monitoring.

树木物种识别与制图对于森林经营、生物多样性保护以及生态学研究均具有至关重要的意义。树皮图像可通过地面拍摄便捷获取，且能提供大量有关树木物种及其健康状况的信息。然而，现有的树皮图像数据集往往范围有限，在物种覆盖多样性与时间属性维度上存在不足。为解决上述局限，本研究提出BarkVisionAI数据集：该数据集涵盖采自印度不同森林类型的13个物种共计167361张树皮图像，内容全面丰富。每张图像均标注有物种名称、设备属性与时间戳，为开展物种识别研究以及树皮特征变异性分析提供了坚实的基础。我们将为每张图像提供详细的元数据信息，以期推动该数据集在生态学研究、机器学习模型训练以及环境监测领域的应用。采用标准图像分类模型开展的基准测试实验验证了该数据集的实用性与有效性，凸显其作为宝贵资源在开发可靠的现实世界自动化树木物种识别与环境变化监测应用方面的潜力。