Friday, May 17, 2019 (put topic here)

Participants: Thomas, Richard, Subir, Michael, Katharina, Josefine, Sophia, Maren, Malin, Ulf and Axel

Topics of today:

  • summary from last week

Preparations for meeting with man from Hermerle next Friday (24.5.)

(he is open for cooperation, we will see if we can make use of it)

presentation for him

(Thomas has further notes to this)

presentation (~10min):

  • intro what our backgrounds are/ cognitive science

  • intro to study project : what we are interested in - methodology —> came to idea to apply it to machine of sophia’s brother

  • classes what we want to sort –> include picture of decision tree which Josefine draw

  • explain our toolbox - state openness to other methods

  • share extraction of pictures, which Thomas already did

questions to ask him: (more questions should be added)

  • what exactly is he doing (how is he adjusting the machine) when he is at Rheine?

  • do more farms have the same machine?

  • what are we not allowed to do with the machine?

  • what is he currently doing, software wise - current methods - on new machines?

  • is the machine bad because of hardware or because of software problem?

  • is he willing to help us to integrate a new software on old machine?

  • is there a prior integrated?

  • what is the bucket full behaviour of the machine?

  • question about demo version - (what didn’t work when Thomas and Sophia went to Rheine last time)

  • ask him if we can get software QV

  • thomas, josefine & Sophia feel responsible to prepare slides

  • Sophia & Michael will hold presentation

  • split up questions ? - distribute it with asana or before the meeting

  • maren takes care of coffee & cookies

  • more precise definition of tasks for the next weeks so that its easier to assign oneself to tasks

    • get algorithmic baseline - extract features manually

      • simple python functions: takes an image and extracts thickness / width, rusty, length, color, curvature …

        • preprocess_images(path, outpath)

        • extract_width(preprocessed_image, num_measurements = 5): return List_of_Width - Sophia

        • horizontal_slices(preprocessed_image, num_measurements) begin_end_matrix - Sophia

        • extract_curvature(preprocessed_image): curvature score - Maren

        • check_rust(preprocessed_image): score_of_certainty - Michael

        • check_violet(preprocessed_image): score_of_certainty - Subir

        • check_flower(preprossed_image): score_of_certainty - Richard, Katha

        • extract_length(preprocessed_image): length - Malin

    • human classification - labelling of data

    • non programming tasks:

      • classification with hirarchical classes

      • working with unbalanced classes

      • semi-supervised learning