An annotated dataset of images of Chinese giant salamanders

TL;DR

This paper introduces a new dataset of annotated images of Chinese giant salamanders to aid conservation and AI research.

Contribution

The novel contribution is a dataset with 1386 images and annotations in multiple formats for deep learning applications.

Findings

The dataset includes 1386 images with 1397 annotated bounding boxes.

Annotations are provided in PASCAL VOC and COCO formats for compatibility with major detection frameworks.

The dataset supports intelligent monitoring and conservation of the critically endangered Chinese giant salamander.

Abstract

The Chinese giant salamander is classified as a Class II protected species in China and is recognized as critically endangered by the International Union for Conservation of Nature (IUCN). Due to their unique behavioral patterns, wild Chinese giant salamanders are primarily nocturnal and inhabit areas characterized by complex terrain, which results in limited detection coverage and significant challenges in observation. Consequently, images of wild Chinese giant salamanders are exceedingly rare, and the scarcity of existing data impedes the advancement and application of deep learning-based object detection models. This study constructs and releases a specialized dataset for Chinese giant salamanders, comprising 1386 images and a total of 1397 annotated bounding boxes. All images represent diverse field scenarios and are meticulously annotated in accordance with YOLO (You Only Look Once) labeling specifications. Annotation files are provided in both PASCAL VOC (Visual Object Classes) and COCO (Common Objects in Context) formats to ensure compatibility with leading detection frameworks, including YOLO v8 and YOLO v11. This dataset aims to offer high-quality, multi-scenario annotated data for research in computer vision and conservation biology, facilitating the training and evaluation of models for intelligent monitoring and species conservation of the Chinese giant salamander, thereby promoting the development of visual recognition technologies for endangered species.

Figures