All posts in Algorithm Spotlight

Style Transfer with StyleThief

Style transfer is a term used for reimagining an image with the style of a given piece of art. Recently, various research groups have proposed different approaches to do style transfer. Generally speaking, there are trade-offs between these different techniques.

For example, in one of our previous spotlights we talked about DeepFilter, where you train a model based on a style, and stylize images almost instantaneously with that trained model. You would train for about a day, and later be able to stylize images rapidly. The biggest issue with this technique is that training wouldn’t always yield the best results. You would then need to train it multiple times, which could easily add up to a few days.

StyleThief works differently from DeepFilter. It takes a long time to train for every sample image, but is more robust and yields better stylized images. It is a trade-off between speed and quality. Read More…

Quickly Building a Face Recognizer

Have you ever wondered how companies like Facebook automatically tag millions of user images?

Or did you find yourself in a situation where you want to automate tagging people in images… perhaps with tens of thousands of images?

Or maybe you just want to build a simple TV show celebrity classifier for your fan site?

Well now there’s a quick and efficient way of doing this today. And it scales seamlessly! Read More…

Censoring Faces Automatically

Privacy issues are a big concern when recording public videos. Professional photographers and public institutions (such as police departments) run into this problem when publishing or releasing public images or footage. Generally speaking, professionals are required to obscure faces when there is a reasonable expectation of privacy and the individual(s) being filmed have not signed a release form. Similarly, many governmental institutions censor faces when releasing video footage, in order to maintain the privacy of those in the video.

Censor face tries to solve these problems by automating the process of censoring faces, efficiently and at scale. Read More…

Automatic Scene Detection

Hundreds of thousands of videos are uploaded each day to Youtube, Facebook, Instagram, Snapchat and other sites. One of the many issues that these services face is the extraction of useful metadata. At Algorithmia, we’ve been automating a lot of this process, enabling rapid auto-tagging and feature detection of videos. But there’s still more work to be done…

For example, if you wanted to put an ad in the middle of a video, as an advertiser you’d probably prefer to show the ad in between scene cuts, where it would be less intrusive. Or perhaps you’re curating a stock footage gallery but are working with multi-scene footage which needs to be split up. In both of these cases, you don’t want to manually scrub through many thousands of videos each day to determine the best insertion or cut-points.

Scene Detection is an algorithm which automates this task at scale, and is now available for you to use via Algorithmia.  Let’s dig into the details of how we detect scene changes in videos… Read More…