Data for: From Noise to Nature: Diffusion Models to Advance Wildlife Detection in Remote Sensing Data

Data and code repository for the manuscript “From Noise to Nature: Diffusion Models to Advance Wildlife Detection in Remote Sensing Data.” Below is the overview of the basic directory structure of this folder: The COCO_Files.zip directory contains .json files in COCO format specifying the filenames of the images used in each experimental subset (e.g. experiment1_train.json). The lora.zip directory contains images and prompts for fine-tuning (lora/images), as well as bounding box annotations (lora/labels) used for color channel analyses and model weights (in .safetensors format) for the LoRA model used in this study. The data.tar file contains images (e.g. data/experiment1/images) and corresponding .json annotation files (e.g. data/experiment1/labels) for each experimental subset described in the manuscript, as well as for the validation and test set held constant across all experiments. To access the data, download and run the following through your terminal. tar -xvf data.tar The zipped_code.zip file contains all scripts for analyses done in the paper, as well as a README with instructions for execution.