SalaPrep: A preprocessing tool for salamander photo-identification workflows

SalaPrep is a lightweight GUI tool developed to automate the rotation and cropping of larval Eurycea spp. (Caudata: Plethodontidae) headshots intended for downstream use in an individual identification software (e.g., Wild-ID). The program segments the salamander from the image background using Meta's Segment Anything Model 2 (SAM 2), body axis is inferred using principal components analysis (PCA) via singular value decomposition (SVD) to enable accurate rotaton, and head orientation (facing left vs. right) is determined by a compact MobileNetV2 classifier trained on 1,200 photos to 82% accuracy. Images are cropped to a standardized frame with snout at one edge and the gill/neck junction at the other. Users can process images singularly or in batches, designate animal orientation, manually crop and rotate images if the program generates imperfect results, and modify the prefix or suffix of processed image file names.