This Zenodo record contains two test datasets (Birds and Littorina) used in the paper: PhenoLearn: A user-friendly Toolkit for Image Annotation and Deep Learning-Based Phenotyping for Biological Datasets Authors: Yichen He, Christopher R. Cooney, Steve Maddock, Gavin H. Thomas PhenoLearn is a graphical and script-based toolkit designed to help biologists annotate and analyse biological images using deep learning. This dataset includes two test cases: one of bird specimen images for semantic segmentation, and another of marine snail (Littorina) images for landmark detection. These datasets are used to demonstrate the PhenoLearn workflow in the accompanying paper. Dataset Structure Bird Dataset train/ — 120 bird specimen images for annotation and model training. pred/ — 100 images for prediction and testing. seg_train.csv — Pixel-wise segmentations (CSV format with RLE or polygon masks). name_file_pred — Filenames corresponding to prediction images. Littorina Dataset train/ — 120 snail images for training landmark prediction models. pred/ — 100 snail images for model testing. pts_train.csv — Ground-truth landmark coordinates for training images. name_file_pred — Prediction image filenames for evaluation. How to Use These Datasets Workflow Instructions (via PhenoLearn) Download the dataset folders. Use PhenoLearn to load seg_train.csv (segmentation) or pts_train.csv (landmark) to view and edit annotations. Train segmentation or landmark prediction models directly via PhenoLearn's training module, or export data for external tools. Use name_file_pred to match predictions with ground-truth for evaluation. See the full tutorial and usage guide in the https://github.com/EchanHe/PhenoLearn.