Image Classification Review Tool

• Use Up/Down arrows to switch between labels
• Use Right arrow to save current label and get a new image

Usage

python start.py [--predictions (default: predictions.json)] [--output (default: output.json)]

Example predictions.json file with raw predictions

{
  "images/12_Happy.jpg": [
    { "label": "happy", "score": 0.9760520458221436 },
    { "label": "surprise", "score": 0.007539073470979929 },
    { "label": "neutral", "score": 0.00606749951839447 },
    { "label": "sad", "score": 0.003563844133168459 },
    { "label": "fear", "score": 0.002640938386321068 }
  ],
  "images/10_Contempt.jpg": [
    { "label": "happy", "score": 0.8121405839920044 },
    { "label": "surprise", "score": 0.057953134179115295 },
    { "label": "neutral", "score": 0.046099040657281876 },
    { "label": "fear", "score": 0.031888220459222794 },
    { "label": "angry", "score": 0.02385363169014454 }
  ]
  ...
}

Example output.json file with reviewed labels.

{
    "neutral": [
        "images/0_Anger.jpg",
        "images/0_Contempt.jpg",
        "images/0_Neutral.jpg",
        "images/0_Surprised.jpg",
        "images/10_Neutral.jpg",
        "images/11_Sad.jpg",
        "images/12_Neutral.jpg",
    ],
    "fear": [
        "images/0_Fear.jpg",
        "images/10_Surprised.jpg",
        "images/13_Disgust.jpg"
    ]
    ...
}

Made with HyperDiv