UAV Thermal Dataset for Wildlife Detection: Focusing on Koala

<b>Dataset Description</b>This dataset contains <b>thermal drone imagery of koalas</b>, collected in dense eucalyptus forest environments in Australia. Images were captured at a resolution of <b>640×512 pixels</b> using drone-mounted thermal sensors, designed to address the challenges of ecological monitoring under canopy cover and low-contrast conditions.The dataset includes <b>4,277 training images</b> and <b>1,252 validation images</b>, annotated across two categories:<b>Koala</b> – confirmed thermal detections of koalas.<b>Non-Koala</b> – other heat signatures (e.g., vegetation, branches, or other non-target animals).Annotations are provided in both <b>YOLO </b><code><strong>.txt</strong></code><b> format</b> and <b>COCO-style </b><code><strong>.json</strong></code><b> format</b>, ensuring compatibility with widely used object detection frameworks such as <b>YOLOv5/YOLOv8, MMDetection, and Detectron2</b>.<br><b>Folder Structure</b><pre>Koala/<br>├── images/ # Thermal drone image frames<br>├── annotations_yolo/ # YOLO-format annotations<br>├── annotations_coco/ # COCO-style JSON annotations<br>├── README_Koala.md # Dataset-specific documentation<br><br><b>Annotation Details</b></pre><b>YOLO Format:</b> <code>[class_id x_center y_center width height]</code> with all coordinates normalized to image size.<b>COCO Format:</b> Standard <code>images</code>, <code>annotations</code>, and <code>categories</code> fields compatible with <b>pycocotools</b> and other libraries.<b>Potential Applications</b><b>Wildlife monitoring and conservation</b> – improving automated detection of koalas for ecological studies.<b>Small object detection in thermal UAV imagery</b> – benchmarking object detection under canopy occlusion.<b>Model generalization studies</b> – evaluating transferability and domain adaptation for thermal datasets.<b>AI for conservation</b> – supporting tools to aid field researchers in real-time koala population surveys.<br>