Arthropodia-V1: Biodiversity Dataset for Semantic Segmentation, Quality Estimation, and Observation Extraction

This paper introduces a novel biodiversity image dataset named Arthropodia, which focuses on three crucial tasks: image quality estimation, semantic segmentation, and observation extraction (image similarity). Arthropodia is collected from tropical Malaysian forests and encompasses a diverse range of arthropod species captured under various lighting and environmental conditions. We manually annotate the dataset for each task, making it suitable for benchmarking and developing AI models applicable to biodiversity and environmental monitoring. To demonstrate its utility, we present baseline results using state-of-the-art deep learning models for each task. Our findings indicate that Arthropodia serves as a valuable resource for advancing research in the field of biodiversity conservation and ecological studies through automated image processing and analysis. The open-access availability of Arthropodia fosters further contributions from the research community, paving the way for the development of robust and accurate AI solutions for biodiversity monitoring and preservation.