This HCCreadme.txt file was generated on 20200225 by ALEXANDRU GRIGOROIU.

-------------------
GENERAL INFORMATION
-------------------

Title of Dataset: Hyperspectral Colour Classification (HCC)

Author Information

	Principal Investigator: Sarah E. Bohndiek, Department of Physics, University of Cambridge, JJ Thomson Avenue, Cambridge CB3 0HE, United Kingdom, seb53@cam.ac.uk
	Associate or Co-investigator: Alexandru Grigoroiu, Department of Physics, University of Cambridge, JJ Thomson Avenue, Cambridge CB3 0HE, United Kingdom, ag745@cam.ac.uk
	Alternate Contact: Jonghee Yoon, Department of Physics, University of Cambridge, JJ Thomson Avenue, Cambridge CB3 0HE, United Kingdom, jy385@cam.ac.uk

Date of data collection: 20170915 to 20190526

Geographic location of data collection: Cambridge, Cambridgeshire, United Kingdom 

Information about funding sources or sponsorship that supported the collection of the data:
Work was funded by the UK Engineering and Physical Sciences Research Council (EPSRC) (EP/N014588/1, EP/R003599/1), CRUK (C47594/A16267, C14303/A17197, C47594/A21102) and the EU FP7 agreement FP7-PEOPLE-2013-CIG-630729. AG was supported by the EPSRC grant for the University of Cambridge Centre for Doctoral Training in Sensor Technologies and Applications (EP/L015889/1).


--------------------------
SHARING/ACCESS INFORMATION
-------------------------- 

Licenses/restrictions placed on the data, or limitations of reuse:
CC BY

Recommended citation for the data:

Citation for and links to publications that cite or use the data:

--------------------
DATA & FILE OVERVIEW
--------------------

Data is sorted into the figure folder towards which it contributed most.
Dataset is formated into .mat -v7.3 files. They have been normalised using white background and dark noise.
Filename structure: f_dataset_number_grating_YYYYMMDD.mat, where f can take values (s[spectrum] - spectrum, t[target] - rgb image)
Additional data:
	- reference_spectrum: contains the spectrometer data for each colour
Algorithms presented in .py OR .m formats.

--------------------------
METHODOLOGICAL INFORMATION
--------------------------

Description of methods used for collection/generation of data:
Images were captured using a line-scanning hyperspectral endoscope as described in 'Yoon, J., Joseph, J., Waterhouse, D.J. et al. A clinically translatable hyperspectral endoscopy (HySE) system for imaging the gastrointestinal tract. Nat Commun 10, 1902 (2019)'

Methods for processing the data: 
Pixel-wise and Patch-based Convolutional Neural Networks (CNN) implemented in Python 3.6.4 using Theano 1.0.0.
Support Vector Machine (SVM) implemented in Python 3.6.4 using sklearn 0.19.2.
Spectral unmixing and Pearson Correlation Analyis implemented in Matlab R2018a.
The NVIDIA GeForce GTX 1060 Graphical Processing Unit (GPU) was used for acceleration of the learning algorithms.

People involved with sample collection, processing, analysis and/or submission: 
A.G. assembled, analysed and interpreted the data and wrote the manuscript. J.Y. performed the data collection. J.Y. and S.E.B conceived and designed the study. S.E.B. wrote the manuscript.