P14: Visualisation of pollen rehydration via deep learning

James Grant-Jacob, Matthew Praeger, Robert Eason, and Ben Mills

University of Southampton, UK

The size and shape of pollen grains can be an indicator for crop health in agriculture1, as the pollen grain structure can vary during development and dispersal, owing to environmental factors2–4. Pollen grains dispersed in the atmosphere can often be found to be smaller than their original size and of a non-uniform shape due to dehydration, and therefore identification of growth conditions from the pollen appearance can be more challenging. Therefore, the capability for transforming an image of a dehydrated pollen grain into an image of its former hydrated state could assist in the identification of the original size, shape and structure of the pollen grain. To date, deep learning has been used for pollen identification in images5 and from scattering patterns6, and owing to deep learning’s capability to perform image-to-image transformations, it has been used for generating pollen grain images from their scattering patterns7. Here, we use a generative deep learning neural network to transform images of dehydrated Ranunculus pollen grains into images of their hydrated forms. We also use a classification neural network to identify the pollen species from generated images of the hydrated Ranunculus pollen grains. This work could be a step towards a universal model for understanding pollen growth under environmental conditions, such as climate change.

[1]      Fernandez-Mensaque, P. C. et al. Aerobiologia (Bologna). 14, 185–190 (1998)

[2]      Firon, N. et al. Ann. Bot. 109, 1201–1214 (2012)

[3]      Pacini, E. et al. Protoplasma 228, 73 (2006)

[4]      Ejsmond, M. J. et al. Ecosphere 2, art117 (2011)

[5]      Khanzhina, N. et al. Comput. Biol. Med. 140, 105064 (2022)

[6]      Grant-Jacob, J. A. et al. J. Phys. Photonics 1, 44004 (2019)

[7]      Grant-Jacob, J. A. et al. Environ. Res. Commun. 2, 075005 (2020)

Key dates

Registration deadline:

31 January 2022

Organised by the IOP Food Physics Group