Campaigns in Bitskout
Distributing volume based tasks and rewarding participation
Campaigns are one of the key concepts in Bitskout. Campaigns are used to distribute a highly repetitive volume tasks like collecting data, labelling data, giving feedback and etc. In this post we want to have a closer look at how a campaign looks like in Bitskout.
A campaign has the following key sections:
While it is clear what to do with General section, let's check each section for more details.
The storage section allows you to specify a storage service for your campaign. Bitskout provides a possibility to integrate various services like S3 or Microsoft OneDrive. Once you create a campaign, the storage service is required to store the submitted content (images, videos, files). Depending on your configuration, the user might be asked to be authenticated to a storage service before submitting any data.
As an option, all submitted content can be encrypted to provide extra protection.
Task Management section
The Task Management integration is not required in Campaigns, but for the sake of integrity we will cover it here.
With Task Management service integration, you can use your current task management system to define tasks, and then use Bitskout infrastructure to add task validation workflows, storage and payment functionalities making project management complete end to end.
A workflow is a combination of validation steps to automate the submission management. Simply, workflows define steps and conditions of how the submission is analysed to be approved or rejected. A workflow step can be of different type - it can be a manual approval (e.g. an admin or a validator has to click on something) or a Machine Learning model with an accuracy threshold.
Workflows are a very powerful tool that allow you to create a logic and define steps of approvals automating validation of real work. Let's take a look at a workflow step of analysing an image submission.
Here we define the parameters of image analysis. Take a closer look at "Model ID" - you can specify here an actual machine learning model used to recognise an image (object detection type). It can be a custom proprietary model or a 3rd party model. The main parameter here is a minimum accuracy (we analysed if there is a specific equipment in the photo), and if it's met, then the we proceed to the next step of analysis:
Using logic connections you can design your own approval workflows to automate content analysis.
The compensation section allows to define a reward per each approved submission. You can choose from the following options:
The compensation section is the one that requires most effort due to various legal and controlling aspects and usually would require a separate integration and stress testing.
In the next post will cover how a campaign looks like from a user perspective.
We just sent you an email. Please click the link in the email to confirm your subscription!