Remember films like Terminator or Iron Man, where there is an artificially intelligent computer that reads and interprets the world for the hero?
The films illustrate the AI seeing things by using a square or a circle that zeros in and marks the object. This is a real-life process known as data annotation, and it’s the first step necessary to enable AI and machine learning (ML) algorithms to identify what they see.
Without properly labeling the data, the machine only sees digital images and videos of the environment – without the ability to understand and interact with it.
“As machine learning becomes increasingly important for virtually every industry, we aim to arm companies with the tools they need to scale data annotation and build their models efficiently and precisely.”
Keymakr CEO and founder Arie Zilberman
Data annotation
An AI/ML algorithm is only as accurate as the data with which it’s working. If the data fed to the algorithm is inaccurate, it will go off in the wrong direction and reinforce itself later on. Imagine an AI that maps the human genome looking for indications of rare diseases. Small deviations could be the indicator of a rare disease, and the cases might be so unusual that there isn’t a large sample from which to learn. Suddenly, the AI is flagging the potential for a rare disease where there shouldn’t be one, or even worse, missing someone who is at a higher risk.
Just as human doctors’ success depends on the education they received, AI can only perform as well as the training data used to create it.
Today, around 80% of the work that goes toward developing AI capabilities lies in collecting and preparing data. Both small and large organizations struggle to process large amounts of data in a cost-efficient way. Quick and affordable tools to improve ML algorithms are sorely needed, and this is the space into which Israel-based start-up Keymakr has ventured.
“As machine learning becomes increasingly important for virtually every industry, we aim to arm companies with the tools they need to scale data annotation and build their models efficiently and precisely,” says Arie Zilberman, CEO and founder of Keymakr.
Keymakr and Keylabs
The company’s Keylabs platform offers companies ranging from large car manufacturers to small data-annotation companies access to Keymakr’s highly precise video-annotation tools.
This empowers companies in industries ranging from automotive to retail and waste management to scale video and image annotation, and thereby their AI algorithms overall.
Founded in 2015, Keymakr offers high-quality and affordable training data for computer vision-based AI. It is already working with companies like Zeekit, which was acquired by Walmart, and many smaller firms. Keymakr also works with retail companies that deal with images and videos that come from mobile devices, in-store cameras and logistics and warehouse cameras.