Data and Resources




This dataset has no data

Additional Info

Field Value
Title Irish Grass Clover Dataset (VistaMilk)
License CC BY-NC-SA
Teagasc Department Animal and Bioscience Research
Teagasc Programme Animal and Grassland
Description High-resolution images of grass-clover swards at Teagasc Moorepark, with ground-truth annotations for dry herbage mass and sward composition. The dataset was collected by researchers in the VistaMilk SFI Research Centre to develop computer vision and machine learning methods for estimating herbage biomass and clover content from images. It has two subsets: (1) Camera and phone images (2020): 0.5 x 0.5 m quadrat images from a tripod-mounted Canon EOS 90D and an iPhone 6, taken across 26 plots (5-6 images per plot). There are 525 labelled camera images (418 train, 107 validation), 124 labelled phone images, 1,072 unlabelled camera images and 1,112 unlabelled phone images. Labels per quadrat: total dry herbage mass (kg DM/ha); dry and fresh grass, clover and weed percentages; and sward height after cutting (cm). (2) Drone images (late autumn 2021): 331 images from a DJI Mavic 2 Pro over 23 paddocks (7-36 images per paddock, 6-12 m altitude, 5472 x 3648 px), with paddock-level dry herbage mass from visual expert estimates and from strip cutting. The images can be used for training and testing supervised, semi-supervised and unsupervised models of pasture biomass and clover estimation.
Language English
Data creator(s)
  1. Albert, P.
  2. Saadeldin, M.
  3. Narayanan, B.
  4. Mac Namee, B.
  5. Hennessy, D.
  6. O'Connor, A.
  7. O'Connor, N.
  8. McGuinness, K.
Geographic coverage Teagasc Moorepark Farm, Fermoy, Co. Cork
Digital Object Identifier (DOI) https://doi.org/10.5281/zenodo.14191859
Citation University College Dublin, & Teagasc - The Irish Agriculture and Food Development Authority. (2024). Irish Grass Clover Dataset (VistaMilk) [Dataset]. Zenodo. https://doi.org/10.5281/zenodo.14191859
Landing page https://zenodo.org/records/14191859
Rights notes https://creativecommons.org/licenses/by-nc-sa/4.0/deed.en
Project funders VistaMilk SFI Research Centre (Science Foundation Ireland, now Research Ireland, and the Department of Agriculture, Food and the Marine; grant SFI/16/RC/3835)
Related resources
  1. Albert, P., Saadeldin, M., Narayanan, B., Mac Namee, B., Hennessy, D., O'Connor, A., O'Connor, N. & McGuinness, K. (2021). Semi-supervised dry herbage mass estimation using automatic data and synthetic images. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops, pp. 1284-1293
  2. Albert, P., Saadeldin, M., Narayanan, B., Mac Namee, B., Hennessy, D., O'Connor, N. E. & McGuinness, K. (2022). Unsupervised domain adaptation and super resolution on drone images for autonomous dry herbage biomass estimation. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, pp. 1636-1646
  3. Albert, P., Saadeldin, M., Narayanan, B., Mac Namee, B., Hennessy, D., O'Connor, A. H., O'Connor, N. E. & McGuinness, K. (2022). Utilizing unsupervised learning to improve sward content prediction and herbage mass estimation. arXiv:2204.09343
Provenance information Camera and phone subset (2020): images were taken over 0.5 x 0.5 m quadrats on 26 grass-clover plots at Moorepark Farm (Teagasc), using a tripod-mounted Canon EOS 90D and a handheld iPhone 6. For labelled quadrats, herbage was cut at 2-4 cm above ground with Gardena Accu 60 hand shears immediately after imaging. Fresh weight was recorded, and the herbage separated into grass, clover and weeds, oven-dried for 16 h and weighed for dry matter. Further unlabelled images were taken at random locations in the same plots. Drone subset (late autumn 2021): 7-36 images per paddock were taken over 23 paddocks with a DJI Mavic 2 Pro at 6-12 m. Paddock-level dry herbage mass came from (a) visual estimation by two experts at the time of imaging and (b) the protocol of Egan et al. (2018): two 1.2 x 8 m strips per paddock cut at 4 cm with an Etesia mower, and a 100 g subsample dried at 95 C for 16 h.
Time of data collection Camera and phone images 2020; drone images late autumn 2021