PyDaemo
Python Wrapper for Daemo.
PyDaemo should allow you to create and manage all your crowdsourcing tasks for Daemo programmatically.
Read the documentations at pydaemo.readthedocs.io
Download and Installation.
To download and install PyDaemo, run the following command:
pip install pydaemo
To install from source, run:
git clone https://github.com/ranjaykrishna/pydaemo.git
cd pydaemo
python setup.py install
Getting started with PyDaemo.
The first step to use the API is to download your credentials so that you have permissions to communicate with Daemo. Download the access tokens as a JSON file by navigating to your Daemo Profile and clicking "Get Credentials" under the menu in your profile.
Ensure that the credentials you obtain contain a client_id, access_token and refresh_token. Create and save your credentials in credentials.json file. Let's explore the API with two use cases. In the first example below, we will use PyDaemo to create, launch and collect surveys filled in by crowd workers.
Let's create a simple survey project.
Let's create a simple survey task with 3 questions:
- A radio button question.
- A text question.
- A dynamic question who's question will be filled in dynamically for each task.
The script to automatically create such a task is located here: tutorials/simple_survey.py.
First, let's intialize the Daemo's api endpoints.
daemo = Daemo(update_credentials=False)
Next let's create a project and a template.
project = daemo.create_project(name='Simple Survey Project', price=0.2,
template_name='Simple Survey Template')
print('created project: ', project)
> created project: {"id": 37, "template_id": 14}
Now, we will add the first question: A radio button where the options are yes and no laid out in a row. We want to ask the question Isn't this survey east?. Let's make sure that this question is a required question.
options = [{'value': 'yes', 'position': 0},
{'value': 'no', 'position': 1}]
required = True
question = 'Isn\'t this survey easy?',
question_subtype = None
question_name = 'Q1'
question_type = 'radio'
predecessor = None
question1 = daemo.create_template_item(question_name, question_type,
question_subtype, predecessor,
required, project['template_id'],
question, layout='row', shuffle=False,
options=options)
print('created question1 with id: ', question1)
> created question1 with id: 10
Next, let's create the question with a text field. We are asking Is there anything else you would like to say?. We are setting this question to be optional (not required).
required = False
question = 'Is there anything else you would like to say?'
question_subtype = 'text'
question_name = 'Q2'
question_type = 'text'
predecessor = question1
question2 = daemo.create_template_item(question_name, question_type,
question_subtype, predecessor,
required, project['template_id'],
question,
placeholder='placeholder')
print('created question2 with id: ', question2)
> created question2 with id: 11
The third question is going to be a dynamic question that will be set from the data passed in. This allows requestors to customize each task. To create such a task, we simply need to set the question to be {{VARIABLE}} and make sure that we create a task with a dictionary containing the header VARIABLE. Here is what it will look like in code:
required = False
question = '{{dynamic_question}}'
question_subtype = 'text'
question_name = 'Q3'
question_type = 'text'
predecessor = question2
question3 = daemo.create_template_item(question_name, question_type,
question_subtype, predecessor,
required, project['template_id'],
question,
placeholder='placeholder')
print('created question3 with id: ', question3)
> created question3 with id: 14
And we can create a task like this:
daemo.create_task(
project['id'],
{'dynamic_question': 'Woah, a dynamic question!'})
Finally, we can publish the project and wait for workers to complete the task:
response = daemo.publish_project(project['id'])
print(response)
> {"message": "Project :projectName successfully published"}
Tutorial: advanced image captioning task.
Now, let's try something a little more complicated. Let's create a task where we ask workers to caption an image. First we need a list of urls:
urls = ['https://cs.stanford.edu/people/rak248/VG_100K/2322397.jpg',
'https://cs.stanford.edu/people/rak248/VG_100K/2322398.jpg',
'https://cs.stanford.edu/people/rak248/VG_100K/2322399.jpg']
We will initialize Daemo and create a project:
daemo = Daemo(update_credentials=True)
project = daemo.create_project(name='Image Captioning Project', price=0.2,
template_name='Image Captioning Template',
verbose=True)
Next, we will create an image field with its src set to {{url}}. We will pass in different urls when creating individual tasks.
question_name = 'image'
question_type = 'image'
question_subtype = None
predecessor = None
required = False
question = 'An image you need to caption.'
image = daemo.create_template_item(question_name, question_type,
question_subtype, predecessor,
required, project['template_id'],
'', src='{{url}}',
verbose=True)
The last item we need to add to the template is a text field where workers will add a caption. This field will be required:
question_name = 'caption'
question_type = 'text'
question_subtype = 'text'
predecessor = image
required = True
question = 'Caption the image.'
placeholder = 'add your caption here...'
caption = daemo.create_template_item(question_name, question_type,
question_subtype, predecessor,
required, project['template_id'],
'', placeholder=placeholder,
Let's create the tasks with the individual urls and then publisht the project.
for url in urls:
daemo.create_task(project['id'], {'url': url}, verbose=True)
daemo.publish_project(project['id'], verbose=True)
Tutorial: getting results and approving work.
Coming soon.
Tutorial: using custom iframes to create tasks.
Coming soon.
Contributing to the repository.
We gladly welcome contributions that improve the API or even provide additional tutorials that demonstrate how to use PyDaemo. Create a fork of this repository and send a pull request.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent