new dataset with just normal and opacity lungs

[kmader]

• 50 cases
• 25 positive
• 25 negative
• questions (would this image ... described as: no opacity present, opacities present)

[betatim]

How about "Is there an opacity present in this image?" and "Is there no opacity present in this image?" with a description of what exactly is meant with "opacity"?

[kmader]

I'll add a field to the dataset in the task.json containing the question to use for each category (otherwise it will just use the default one). As I guess a lot of datasets will have 'normal' for a category this seems like it should be useful.

import json
annotation_task = {
    'google_forms': {'form_url': 'https://docs.google.com/forms/d/e/1FAIpQLSemGag9uitnPnV6OBHDNgrvr2nh-jArJZhVco0Kfjkx4eRkYA/viewform', 
    'sheet_url': 'https://docs.google.com/spreadsheets/d/1JUCLX_17JIGit0Nk4wphgTHlmji9u9PYPmyf_9Wscvg/edit?usp=sharing'
            },
    'dataset': {
        'image_path': 'Image Index', # column name
        'output_labels': 'class', # column name
        'dataframe': dataset_df.drop('age_group', 1).to_dict(),
        'base_image_directory': 'sample_data', # path
        'questions': {'Normal': 'Is there no opacity present in this image?', 
                      'Lung Opacity': 'Is there an opacity present in this image?'} 
    }
}

with open('task.json', 'w') as f:
    json.dump(annotation_task, f, indent=4, sort_keys=True)