Data for: Deep-Learning based StomaDetector for Maize Leaf Imprints

We took images of maize leaf imprints with an OLYMPUS BX61 microscope at a magnification of 10x (final image resolution: 2448x1920). The training-validation-test dataset (data.zip) consists of a diverse set of 250 maize leaf imprints at developmental stage V6. All stomata in these images were labelled using bounding box information, resulting in a total of > 10'000 objects. This dataset was used to train a deep-learning based object detection model to automatically detect stomata in such microscopy images (included in models.zip). Finally, we evaluated the correlation between our model and ground truth calculations of stomatal density in 2 different lines of maize. For reference, the predicted stomatal counts are saved in predictions.zip and the ground truth data of the inference set is saved in inference.zip. The Images of these 2 lines (NILs and ZmAbh mutants) are available in their respective zip files.

The accompanying code for this dataset is available under: https://github.com/grimmlab/StomaDet