Focus and Motion Blur in Microscopy (FaMM)

Citation Author(s):
Patrick
Krawczyk
Materials Research Institute Aalen, Aalen University, Beethovenstrasse 1, D-73430 Aalen Germany
Andreas
Jansche
Materials Research Institute Aalen, Aalen University, Beethovenstrasse 1, D-73430 Aalen Germany
Timo
Bernthaler
Materials Research Institute Aalen, Aalen University, Beethovenstrasse 1, D-73430 Aalen Germany
Gerhard
Schneider
Materials Research Institute Aalen, Aalen University, Beethovenstrasse 1, D-73430 Aalen Germany
Submitted by:
Patrick Krawczyk
Last updated:
Mon, 07/08/2024 - 15:58
DOI:
10.21227/vwrx-yw83
Data Format:
*.png
Links:
Materials Research Institute, Aalen University
License:
Creative Commons Attribution
154 Views
Categories:
Artificial Intelligence
Computer Vision
Keywords:
Image deblurring, Material Imaging, light-optical microscopy
0
0 ratings - Please login to submit your rating.
ACCESS DATASET CITE SHARE/EMBED

Abstract 

A novel image deblurring dataset for materials science and light-optical microscopy. This dataset provides images with real out-of-focus and motion blur and a sharp reference image for each observation. The dataset includes image samples of lithium-ion batteries, Fe-Nd-B sintered magnets, 100Cr6 steel with a partially bainitic microstructure, and aluminium-silicon casting alloys. The dataset was acquired using a ZEISS AxioImager.Z2 Vario light microscope and the 6-megapixel camera Axiocam 506 color. The dataset includes a total number of 35709 acquired images and was split in the ratio of 8:1:1 for train, validation, and test, ensuring that each material class was equally distributed.

Instructions: 

The dataset is already divided into train, validation, and test.

Abbreviation of materials

  • LIB = lithium-ion battery
  • Magnet = Fe-Nd-B sintered magnet
  • Bainit = 100Cr6 steel with a partially bainitic microstructure
  • AlSi = Aluminium-silicon casting alloy

The file name convention of the images is as follows:

  • date_material_magnification times 10_Y coordinate_X coordinate_mode_motion_focusdeviation in micrometer
  • e.g. the image 20200724_LIB4_20.0_10_1_Focus_0_1.779.PNG
    • was acquired on the 24.07.2020
    • is a lithium-ion battery sample (LIB)
    • was acquired with a magnification of 200x
    • Y-coordinate = 10
    • X-coordinate = 1
    • mode = Focus
    • motion = 0
    • focus deviation of 1.779 micrometers

Getting an image pair of a blurred image as input and sharp reference image as label

  1. The sharp reference image and the blurred image are in the same folder
  2. Sharp reference images contain "ZStack" in the file name e.g. 20200724_LIB4_20.0_10_1_ZStack_0_253.42.PNG
  3. Blurred images contain Focus, Vibration or Both as mode in the filename e.g. 20200724_LIB4_20.0_10_1_Focus_0_1.779.PNG
  4. Make sure the date, material, magnification and Y- and X-coordinates in the file names match e.g.
  • 20200724_LIB4_20.0_10_1_Focus_0_1.779.PNG  = blurred input image
  • 20200724_LIB4_20.0_10_1_ZStack_0_253.42.PNG  = sharp reference image
  • (See dataset files for an example of a data class in PyTorch)

    •  

     

    Funding Agency: 
    Federal Ministry of Education and Research of Germany
    Grant Number: 
    13FH176PX8

    Dataset Files

    LOGIN TO ACCESS DATASET FILES

     

    QUESTIONS?