Friday, June 21, 2019 (Filename information(!), Semi-supervised Learning, Data Augmentation)

Attendants: Axel, Richard, Malin, Luana, Maren, Josefine, Katharina

Outline for the session

  1. reporting about Continuos Integration with Travis from Katha (see travis.yml for more info)

  2. presentation from Luana (Data Augmentation, Super Resolution)

  3. presentation from Josefine (Semi-supervised learning)

  4. report about visit to Querdel

Data collection

  • still problems to catagorize pictures from the sorted collection –> look at the timestamp in the filename

  • Example: G01-190411-092055-873_F00.bmp

    • This picture is the first (F00) of three images (F00, F01, F02) (do not rely on this!)

    • G01: (?)

    • 190411: 11th April 2019

    • 092055: 09:20:55 (nine o’clock and 20 min and 55 sec)

    • 873: constant number for one asparagus piece

Semi-supervised learning

Link to presentation: https://tinyurl.com/yydup2b7

Data Augmentation and Single Image Super-Resolution

Link to presentation: https://tinyurl.com/y2mjm49b (Short description of data augmentation methods, and overview of current SISR algorithms)

Visit to Querdel

  • 6000 images unlabeled (few category flower, category normal)

  • 5 persons working on the machine (3 people resorting)

  • quality of head images is much better

  • the light set up is better (less reflections)

  • still not sure if we can make use of the images