multi_image_classification.py

# coding: utf8
from __future__ import unicode_literals

import prodigy
from prodigy.components.loaders import Images
from prodigy.util import split_string
from prodigy.util import set_hashes
from prodigy.components.db import connect
import itertools
                
def update(answers):
    for answer in answers:
        for option in answer['options']:
            del option['image']
        # print('answer:', answer)

# Recipe decorator with argument annotations: (description, argument type,
# shortcut, type / converter function called on value before it's passed to
# the function). Descriptions are also shown when typing --help.
@prodigy.recipe('multi_image_classification',
    dataset=("The dataset to use", "positional", None, str),
    source=("Path to a directory of images", "positional", None, str),
)
def multi_image_classification(dataset, source):
        
    tmp = Images(source)
    stream = []
    it = 0
    tID = 1
    x = {}
    for task in tmp:
        if it == 0:
            x = {}
            x['label'] = 'DIY'
            x['id'] = tID
            x['options'] = []
        x['options'].append({'id': it, 'text': task['meta']['file'] ,'image': task['image']})
        if it == 3:
            stream.append(x)
            it = 0
            tID += 1
        else:
            it += 1

    if it > 0:
        stream.append(x)

    return {
        'view_id': 'choice', # Annotation interface to use
        'dataset': dataset,        # Name of dataset to save annotations
        'update': update,
        'stream': stream,          # Incoming stream of examples
        'config': {             # Additional config settings
            'choice_style': 'multiple',
            'choice_auto_accept': False
        }
    }