mushon / The-Normalizing-Machine

An experiment in machine learning & algorithmic prejudice
http://mushon.com/tnm
1 stars 0 forks source link

implement filesystem sync between input and output apps #132

Open mushon opened 5 years ago

mushon commented 5 years ago

dataset: https://github.com/dstavy/TNM2/blob/master/bin/data/records/dataset.json

faces (for debugging and kickstarting purposes, do not push these into Github): https://www.dropbox.com/s/r5a0c6bqudewj0x/Faces.zip?dl=0

Input should choose 4 ids to choose from: The latest participant + 3 random (from those who have participated the least)

Round 1: A (latest participant) + B

Round 2: C + D

Round 3: B + C (round 1+2 losers)

Round 4: A + D (round 1+2 winners)

Round 5: A + User (winner vs. user)

The data should be saved into dataset.json during the winner screen display and include:

  {
    "id": "berlin_ars_1",
    "gender": "female",
    "age": 39,
    "beard": false,
    "hairColor": "gray",
    "glasses": 0.8,
    "score": 0.3,
    "selections": [
      "berlin_ars_5",
      "berlin_ars_2",
      "berlin_ars_3",
      "berlin_ars_4"
    ],
    "timestamp": 1567243546
  }

Additional save into backup.json (no score):

    "id": "berlin_ars_1",
    "gender": "female",
    "age": 39,
    "beard": false,
    "hairColor": "gray",
    "glasses": 0.8,
    "selections": [
      "berlin_ars_5",
      "berlin_ars_2",
      "berlin_ars_3",
      "berlin_ars_4"
    ],
    "timestamp": 1567243546
  }

The new participant's image will be recorded during the intro screen (before the interaction) and will be added to the Faces folder.

The update of the dataset will trigger the output screens. @dstavy @eranws please edit if I missed anything