The ability to recognize the semantic content of photographs and video is very important for Adobe. People are the most common and the most important targets of photographers. Knowing the locations of people can lead to more intelligent image and video organization, editing and search tools.
We have developed the Soft Cascade — a novel method for training object detectors. It is a generalization of the Viola-Jones cascade architecture which allows for training detectors that are faster, more accurate, more compact and conceptually simpler. Our face detector achieves state-of-the-art speed and accuracy.
We are using the Soft Cascade in Photoshop Elements’ Face Tagging feature which extracts faces from consumer’s photographs and provides a quick and intuitive way of tagging them. The face tagging feature has received strong reviews . We are also using the Soft Cascade in the Automatic Red Eye Correction feature in Photoshop Elements.