2 mins read
Fascinating: Crowdsourcing, a Man and a Prize
Crowdsourcing is extraordinary because it incorporates parallel thinking and imagination at a vast scale, by engaging a crowd of people for problem-solving situations to transform the world.
As for its origin, did you know soda gave us crowdsourcing?
The reward was 2,400 livres (around 180 pounds sterling around the time). The problem was: How to produce sodium carbonate (soda) from sodium sulfate (sea salt) at an industrial scale? To make soap, glass, dyes, and bleaches at a low-cost and acceptable level of purity?
This example of crowdsourcing goes back to the 18th and 19th centuries. In an era where soap and glass were produced using potash extracted from wood ash. The French Academia of Science, by request of The French King Luis XVI, launched a prize to find a solution, and Nicolas Leblanc found it.
In a few words, crowdsourcing is the concept of outsourcing using a crowd in a collaborative way. It could be internet-based (online-based) to co-creation, packaging, or innovating projects and ideas.
The practice of crowdsourcing existed beyond Jeff Howe’s famed Wired article “The Rise of Crowdsourcing,” in June 2006. Yet, he coined the term.
Beyond the outstanding contribution of chemical engineers such as Nicolas Leblanc. And how his process “for some 60 years in the middle 19th century has influenced European chemicals industries” until today.
What is crowdsourcing in machine learning and how does it work?
Image annotation techniques such as semantic segmentation, bounding box, 3D bounding box, and emotion recognition are key for computer vision models to understand objects’ location, shape, and direction.
Use cases include sensory robotics applied to help with waste management and manufacturing 4.0. AMP Robotics object’s sorting systems use image sensors, controllers, and pushing mechanisms to process images and then sense and locate targeted objects.
“Pick and Place Systems” make manufacturing workplaces safer for employees. The pick and place robots help to lower the rate of fatal workplace injuries.
But to keep up to date, these “recycling robots” or “sorting robots” need ample image training data to enhance speed and accuracy for their computer vision models. Hence, they require continuous leverage of quality image data and image data annotation expertise.
Using crowdsourced data for machine learning means utilizing crowdsourcing companies or crowdsourcing platforms for data collection or data labeling services. Often, crowdsourcing companies have a set process custom to machine learning project requirements such as:
|✅ The client submits the project requirements and uploads the raw data such as audio, image, text, or video; to the in-house platform.|
|✅ The project is supervised and reviewed by project and resource managers in charge of assessing the number of data annotators who match the criteria for the projects.|
|✅ The data labelers are organized and asses with a set quality assurance process following the agreed quality requirements.|
|✅ Data labelers start labeling, executing, and delivering on time.|
|✅ Labeled data is delivered with a quality assurance reviewed — label by label.|
We would love to know your thoughts. Please comment below.