LeafMachine2 Data: Leaf Outlines for Margin Curvature

This dataset is a small subset of a larger LeafMachine2 dataset for this publication: Weaver, W., Berv, J., and Smith, S.A. (In Prep). Evolutionary History Structures Global Leaf-Climate Relationships. This dataset includes leaf outline masks for the following families: Family Leaf Outlines Salicaceae 151,449 Sapindaceae 118,413 Lauraceae 69,125 Fagaceae 60,857 Moraceae 59,601 Betulaceae 29,821 Viburnaceae 22,229 Ulmaceae 14,306 Cornaceae 9,706 Total 535,507 Each txt file is named like this: MICH_1850441029_Achariaceae_Caloncoba_crepiniana__L__1780-918-2648-2037.txt cooresponding to HerbariumCode_GBIFid_Family_Genus_specificEpithet__L__bboxLocation.txt "MICH_1850441029_Achariaceae_Caloncoba_crepiniana": This is the name of the parent herbarium specimen. "__L__1780-918-2648-2037": These are the bounding box coordinates of the leaf in the parent specimen image and act as a unique identifier for the leaf. Each .txt file has this data format: Line Description 1 Height of the original herbarium sheet image (pixels) 2 Width of the original herbarium sheet image (pixels) 3 Conversion factor — pixels per cm, from scale bar 4 Scaling factor — max extent of the contour bounding box, used to normalize the contour 5 X-offset — minimum x-value before normalization, needed to reverse scaling 6 Y-offset — minimum y-value before normalization, needed to reverse scaling 7 Rotation angle (degrees) applied to align the leaf vertically (tip up, base down) 8 Topmost point of the rotated contour (x,y) 9 Bottommost point of the rotated contour (x,y) 10 Leaf tip — contour point closest to the top after rotation (x,y) — machine learning prediction, more robust than Line8 at determining the lamina tip 11 Leaf base / petiole attachment — contour point closest to the bottom after rotation (x,y) — machine learning prediction, more robust than Line9 at determining the lamina base 12+ Leaf contour coordinates, one x,y point per line, normalized to fit in unit circle We include plot_simple_label.py to show how to access the data. The image AK_1897818999_Achariaceae_Pangium_edule__L__257-580-1276-3024.png shows components labeled and reconstructed from the .txt file AK_1897818999_Achariaceae_Pangium_edule__L__257-580-1276-3024.txt. Since the leaf masks come from herbarium specimens that vary in original resultion (2 megapixels - 100 megapixels) the number of points in the leaf outlines is highly variable.