Digital Herbaria: Leaves
Collection
A curated collection of plant images for a research-based art practice ; connections to environmental humanities and the history of science. • 18 items • Updated • 1
Datasets:
Damaged Leaf Herbarium
An Initial Curated Selection of Insect-Damaged, Regenerated, and Computationally Remediated Leaves
Jennifer L. Karson | Plant Machine Design Group | University of Vermont
📦 Archival dataset record and DOI: https://doi.org/10.7910/DVN/ABDI7R
Description
This is an initial curated release of approximately 1,600 tree leaves selected from the Damaged Leaf Herbarium, a collection of 10,000+ damaged leaves collected by the author in Colchester, Vermont, USA during Lymantria dispar outbreak years 2021 and 2022. The leaves were photographed by the University of Vermont Plant Machine Design Group. Also included is a sample of second generation regrowth — second flush leaves from a Quercus rubra that were initially defoliated in the early spring of the outbreak years. Also included is a sample of damaged leaves remediated through a speculative computational design project that imagines sympoietic plant-machine relationships using a customized version of CycleGAN, one that reveals the visual output of each layer of the neural network.
A full dataset release is planned in conjunction with forthcoming publications and art exhibitions. The larger project explores interspecies entanglements in relation to these insect outbreaks and the historic introduction of Lymantria dispar to North America, and its connection to late 19th century ambitions to establish a silk industry in the United States.
Researchers wishing to be notified of the full release may contact the author at jennifer.karson@uvm.edu.
Dataset Structure
The dataset contains three categories of images:
damaged/ — Original field photographs of Lymantria dispar damaged Quercus rubra leaves, organized into four curated collections (table_i through table_iv) reflecting their original presentation as exhibited artworks. The groupings are curatorial rather than taxonomic.
second_generation/ — Second flush leaves from Quercus rubra trees that were initially defoliated during the outbreak years. Filenames retain their original table prefix indicating curatorial provenance.
remediated/ — Four damaged leaves computationally remediated using a customized version of CycleGAN, organized by leaf ID (leaf_reg_4409, leaf_reg_4415, leaf_reg_4416, leaf_reg_7058). Each leaf folder contains eight images revealing the visual output of each layer of the neural network, from the original damaged source (register_00) to the fully remediated output (register_07).
Metadata
A complete metadata CSV (damaged_leaf_herbarium_V1_metadata.csv) is included with the following fields:
file_name | relative_path | file_type | category | table | leaf_id | register | register_label | collection_years | photograph_date | image_width | image_height | file_size | creator | copyright | rights | license | dataset | dataset_version | source | contact
All images have provenance metadata embedded directly in their EXIF/XMP fields.
Citation
If you use this dataset please cite:
Karson, Jennifer L. Damaged Leaf Herbarium: An Initial Curated Selection of Insect-Damaged, Regenerated, and Computationally Remediated Leaves. Harvard Dataverse, V1, 2026. https://doi.org/10.7910/DVN/ABDI7R
License
This dataset is released under CC BY-NC 4.0. Use requires attribution to Jennifer L. Karson and the Plant Machine Design Group, University of Vermont. Commercial use is not permitted.
Contact
Jennifer L. Karson Plant Machine Design Group University of Vermont jennifer.karson@uvm.edu
- Downloads last month
- 14