How Visual Search Helps Finding the right Video Faster
As MotionElements has announced in a press release, they have just managed to integrate a Visual Search platform for stock video with Pixolution into their website. This is a big technological step towards offering buyers motion content based on the visual appearance instead of the often needed artificial language based keyword systems.
A basic video provided by MotionElements explains how Visual Search works on their site:
We had a chance to look and test it for ourselves. The results are very promising and we are interested to see how it will further develop in the future.
Image based searches are developing quickly
A few years ago, TinEye was one of the first to offer a brand new service – finding images based on an image as a reference. Up till then, the image industry relied heavily on keywording of images as the only way to find content matching any user’s needs.
With the rise of digital cameras, the creation of images exploded, and today’s world is shooting and filming billions of photos and videos each day. Most of the content is not keyworded at all, some can be identified by other metadata like included GPS locations. But very often a search for content starts with a base picture someone has seen somewhere. In those cases, the searcher had to identify the best keywords to describe the base content to find similar images.
After TinEye’s success, other names have joined this field of making the search process for images easier and better: Google Image has integrated an image based search in their own Google Images Site as well.
While TinEye and Google Images are relying on images they just found around the internet – only loosely connected with some text around them -, content agencies also have additional data at hand: Title, Description, Keywords, Author, sometimes GPS data. The integration of all these elements can help the customers to get to the perfect content for their needs far quicker.
Pixolution helps content agencies to integrate visual search features
This kind of feature is also becoming very appealing for image agencies whose business is to deliver the right images for a client’s needs as quickly as possible. Berlin based Pixolution has integrated advanced search features in cooperation with the University of Applied Sciences (HTW) and successfully integrated features that involve both visual search as well as using meta data (keywords and image titles) for several image agencies, among them microstock agency Fotolia, macrostock agency F1 Online and Yuri Arcurs’ PeopleImages. As Pixolution’s Executive Partner and CFO Michael Rost told us:
“pixolution smartly combines the keywords of an image with its visual appearance to search for similar images. As a result not only visually similar images but automatically also images in proper context are presented to the audience. So images or videos that would not show up in a usual result set may appear within the first results – just because they are relevant.”
With Visual Search you don’t need to learn a new language
One of the biggest advantages of Visual Search is its independence from language. Global exchange requires us to think, write and talk in other languages different from our mother tongue. Finding the right words to describe the right content that a computer can understand makes this task often complicated. Even using the language you know by ear is not guaranteed because you need to use the words that a computer will understand and hope that the contributing artist has used the same term to keyword his content.
With Visual Search, language barriers are potentially overcome easily by instead giving the computer more information through images of the kind of content a customer is looking for. With an image, it only takes a few seconds to upload and the search engine will return visually similar results that otherwise might have took longer more time to use the right combination of keywords.
How MotionElements integrates visual search for stock video clips
Now MotionElements is going one step further by using Pixolution’s technology to offer Visual Search not to find images but instead motion content such as footage & animation.
Starting with an image that can be uploaded to the MotionElements website, or based on a video the customer already likes, the Visual Search now finds optically similar videos on the site. The usage of this feature will be interesting to follow as the visual search returns sometimes surprising similarities between videos you would not easily find – as you can see from the first screen shot when the base image of a cloud scape also returns a blossoming cherry tree.
A potential good use for this feature is a client who is looking to cut a sequence of videos and might be looking for smooth visual blending between several short clips – upload a screenshot of how the first clip ends, and you might find the right video to visually continue telling your story.
How to use the Visual Search at MotionElements
Let us check out how this visual search works, it’s pretty easy and straight forward.
- Let’s check out some clips about the “Beach” (you can use any other keyword as well)
- Your result will show some nice beach clips – now click on the loupe icon to start your visual search
- After clicking the loupe icon (or upload your own image or screen-grab on the left) you will see the VisualSearch results showing similar video clips and animations.
- Thats it – see how easy it is to find similar looking clips and animations at MotionElements.
Final Thoughts – Visual Search is the Future!
This will become more interesting opportunities for the future, most likely on both ends: The contributing producers might become less responsible for matching the best keywords to their files, and the clients will no longer have to guess the – sometimes counter intuitive – keywords they would need to find the content they are looking for.
We believe the importance of searching for content visually and/or with other attached information will rise with the amount of content being made available constantly growing. Also intercultural and interlanguage can be overcome with visual searches rather than textual based indexing. In this aspect, we believe MotionElements has done an important step forward, integrating this future technology now. We are looking forward to watch how it develops and is being adopted by clients over time.