CrowdFlower announces a new computer vision solution

CrowdFlower, the AI platform for data science teams, announced at the Train AI conference today enhanced capabilities for their Computer Vision solution designed to simplify and speed up the process of annotating images. As a result, businesses can now annotate millions of images in days and weeks, rather than months and years.

Using CrowdFlower’s Image Annotation for Computer Vision, an image can be labeled down to the pixel, such as footage from drones, planes, or satellites.

CrowdFlower’s platform enables the categorizing, labeling and cleansing of data at scale. As AI systems increasingly enter the mainstream, their usefulness is often defined by the quality of the training data used. While a machine can process complex mathematical equations or structured data in milliseconds, it struggles to make decisions when data is more abstract or subjective, such as images or video. As a result, a machine must be taught and the lesson plan is built using training data.

“Now, new customers come to us and ask us how they should be labeling images for self-driving cars, or consumer product companies identifying SKUs on retail shelves, or aerial images from drones or satellites, or even cancer cells for medical research. We’ve taken the best practices from working with these industry pioneers to drive that learning back into the platform and how we structure successful customer engagements.”, said the CEO of CrowdFlower.

CrowdFlower’s Computer Vision offering allows data scientists to easily create and execute an image annotation job on the CrowdFlower platform.  The platform supports line annotation, bounding box annotation, and pixel-level annotation, and leverages a combination of human and machine intelligence to create large-scale high-quality image training sets.

“A human-in-the-loop workflow is incredibly important for Computer Vision. With something like autonomous driving, errors in the AI system can be devastating,” said Dr. Barney Pell, Ph.D., AI pioneer and Machine Learning Fellow at the Creative Disruption Lab at the University of Toronto. “The quality of modern AI systems based on machine learning is most directly impacted by the quality and quantity of data. CrowdFlower helps companies improve their machine learning models by continuously providing high-quality training data at scales required for performance. Moreover, training a model is not a one-time thing — new edge cases constantly emerge, and that model needs to learn, otherwise it could fail. This is where humans come in to augment this process — serving as a great example of the theory that humans and machines are better together.”

Source: CrowdFlower