Dataset for "Deep learning enabled real time speckle recognition and hyperspectral imaging using a multimode fiber array"
DOI:10.5258/SOTON/D0942

Readme author: Peter Wiecha, University of Southampton
 
Supports the paper
Kürüm, U., Wiecha, P.R., French, R., & Muskens, O. L. (2018). Deep learning enabled real time speckle recognition and hyperspectral imaging using a multimode fiber array. Optics Express, XX(XX), XXXXX-XXXXX. DOI: XXXXX



----------------------------
raw_speckle_data.zip:
- training_speckles_998fibers_43wl.h5
hdf5 file containing the speckle patterns and fiber-core positions on the CCD, used for training of the neural networks for all benchmarks in the paper.
 "speckles": speckle patterns for 2604 fiber-cores at 43 wavelengths between 600 and 700nm. 
 "wl": wavelengths corresponding to the according speckle patterns
 "fiber_pos": position ([x,y] in units of pixels on the raw CCD images) of the fiber-cores on the CCD


----------------------------
figures.zip:
Figure 3:
- fig3a_smiley_raw_data.h5
contains all images shown in figure 3a as arrays of float, representing the greyscale pixel intensity information

- fig3bd_all_data_fullfiber_DL_CS_PI_nonoise.txt
contains column wise and whitespace separated all cross-correlation data, shown in figure 3b and d. Specific information is found in the header of the file.

- fig3c_RGB_raw_data.h5
contains all images shown in figure 3c as arrays of float, representing the greyscale pixel intensity information. Contains furthermore all data shown in figures 10 and 11 of the appendix


Figure 4:
individual .txt files for all shown spectra. Every txt file contains column-wise and in following order: the wavelengths (in nm); the ground truth spectrum; the DL, TR, CS reconstructed spectra.


Figure 5:
- Fig5a_benchmark_CS_DL_TR_noise.txt
contains column wise and whitespace separated all cross-correlation data, shown in figure 5a. Specific information is found in the header of the file.

- Fig5c_benchmark_CS_DL_TR_shift.txt
contains column wise and whitespace separated all cross-correlation data, shown in figure 5c. Specific information is found in the header of the file.


Figure 6:
- visualization2_training_speckles_2604fibers_19wl.h5
hdf5 file containing the speckle patterns and fiber-core positions on the CCD, used for training of the neural networks.
 "speckles": speckle patterns for 2604 fiber-cores at 19 wavelengths between 600 and 700nm. 
 "wl": wavelengths corresponding to the according speckle patterns
 "fiber_pos": position ([x,y] in units of pixels on the raw CCD images) of the fiber-cores on the CCD

- visualization2_speckles_2604fibers_200frames.h5
hdf5 file with the extracted speckle patterns for 200 frames of "la linea" containing several wavelength jumps and mixed wavelength states.



Appendix:

Figure 7:
- training_val_loss_vs_epoch.txt
contains raw data for the validation loss vs neural network training epoch

Figure 9:
- smiley_all_channels_raw_data.h5
contains all images shown in figure 9 as arrays of float, representing the greyscale pixel intensity information

Figures 10, 11:
raw data can be found in the file "fig3c_RGB_raw_data.h5" (see Figure 3)