Shutterstock’s ongoing efforts to make the creative life of its customers easier continues with the launch of
reverse image search for video. This new feature makes it faster than ever to find exactly the right video clip in Shutterstock’s footage library.
This new feature makes the discovery process more intuitive and efficient. It’s rooted in Shutterstock’s history of
machine learning experimentation and
reverse image search innovation. And we powered it with our custom-built convolutional neural network that lets you execute a search by simply uploading an image instead of typing keywords.
Essentially, the “what” and the “how” share the same answer: machine learning. Our neural network studies our library of over 280 million images and more than 15 million video clips. It breaks them down at the pixel level into their principal features. It recognizes what’s inside each and every image — including shapes, colors, and the smallest of details. With every image broken into pixel data, our visually similar search tool offers unparalleled image matching and discovery.
When we rolled out reverse image search for still images in 2016, users loved it. They were excited to have a fast, easy alternative to traditional keyword searches that lets them upload an image to find a similar image. In fact, we’ve found that users of reverse image search are over three times more likely to find what they’re looking for than users of keyword searches. That’s why we’re adapting this powerful search technology to work with the unique aspects of video.
How Does Reverse Image Search for Video Work?
Our neural network analyzed every video clip in our library to identify the video preview image and capture the characteristics of the clip. The result: Footage users can upload a photo, frame, or screenshot that illustrates the clip they need. Then, our algorithm provides videos similar in look and feel to the uploaded image.
This is a wonderfully helpful tool for videographers and video editors, as it makes missing shots a thing of the past. Need footage resembling a certain location? Need a scene that features a certain light temperature, wall color, or object? Just show our neural network what you need and reverse search will do the rest.