Readme for "Dataset for: Lensless imaging of pollen grains at three-wavelength using deep learning"

DOI: https://doi.org/10.5258/SOTON/D1354

README AUTHOR: James Grant-Jacob, University of Southampton ORCID 0000-0002-4270-4247

PAPER:

Grant-Jacob, J., Praeger, M., Loxham, M., Eason, R. W., & Mills, B. (2020). Lensless imaging of pollen grains at three-wavelength using deep learning. Environmental Research Communications. DOI: 10.1088/2515-7620/aba6d1

Figure 1 is the schematic of the experimental setup. Showing (a) Illustration of setup for in-situ optical imaging of pollen grains and collection of their scattered light when illuminated by red, green and blue laser sourceslight. (b) Experimentally measured scattering patterns and associated in-situ optical images of Narcissus and Iva xanthiifolia pollen grains.

Figure 2 is the schematic of neural network training for generating images of pollen grains from their experimental scattering patterns via training that used pairs of experimental images and experimental scattering patterns. This procedure was followed for both in-situ and ex-situ neural network training.

Figure 3 is the capability of the trained neural network to generate images of pollen grains from their scattering patterns, showing the scattering pattern (column 1), the generated image (column 2) and the actual image (column 3), for (a) Narcissus, (b) Populus deltoides (c) Iva xanthiifolia and (d) Mahonia aquifolium. Column 4 displays a comparison metric, where black is true negative, white is true positive, blue is false negative and green is false positive, obtained via thresholding and comparison of the generated and experimental images.

Figure 4 is the capability of the trained neural networks for generating ex-situ images of pollen grains from their experimental scattering patterns (column 1), showing the neural network generated images (column 2) and the ex-situ experimental images (column 3). Row (a) shows results for Hyacinthus orientalis and row (b) Chamelaucium. (i) and (iii) show the z-stack type images whilst (ii) and (iv) show SEM type images (for a side-by-side comparison of generated and experimental images). Sap is also labelled on some of the images. 

Figure 5 is a comparison of the capability of trained neural networks for generating an ex-situ experimental image of a pollen grain from its scattering pattern, showing generated images for Chamelaucium, from neural networks trained with (a) red light, (b) green light, (c) blue light, and (d) all three wavelengths, with the corresponding scattering patterns shown in the inset of each image. For reference, the experimental z-stack image of the pollen grain is shown in (e).

Table 1 contains the percentage of pixels in generated images shown in figure 3, for true negative (black), false negative (blue), false positive (green) and true positive (white).

Projects: BM was supported by an EPSRC Early Career Fellowship (EP/N03368X/1). ML was supported by a BBSRC Future Leader Fellowship (BB/P011365/1) and a Senior Research Fellowship from the National Institute for Health Research Southampton Biomedical Research Centre. 

Licence: CC BY

Readme written July, 2020