HP Labs

HP Labs research underpins HP’s Pixel Intelligence image analysis portfolio

By Simon Firth, HP Labs Correspondent — June 29, 2017

From left: Arjun Patel, member of the HP Print Software Platform team, and Dr. Qian Lin, Distinguished Technologist.

From left: Arjun Patel, member of the HP Print Software Platform team, and Dr. Qian Lin, Distinguished Technologist.


HP has made a powerful portfolio of computer vision algorithms available to companies looking to turbocharge their ability to make sense of visual data.

Face detection: Identify and locate human faces in digital imagery.

Face detection: Identify and locate human faces in digital imagery.

 The HP Pixel Intelligence portfolio unlocks a range of capabilities - from accurately locating, analyzing, and grouping faces and objects to automating image layout and cropping.  HP has already used Pixel Intelligence in some of its own leading imaging products. Now other companies can add these same capabilities to their own products and services.

The algorithms are the fruit of a long running HP Labs research program focusing on computer vision, says Dr. Qian Lin, Distinguished Technologist for Computer Vision and Deep Learning Research in HP’s Emerging Compute Lab.

“We’ve been incrementally improving computer vision for a long time,” Dr. Lin notes. “But in the last few years we’ve been applying deep learning, which combines machine learning with artificial neural networks, to the problem and that has allowed us to make enormous strides in the quality of our results.”

The algorithms in the Pixel Intelligence portfolio can find faces within an image or find the same face in multiple images with great accuracy. They can also recognize specific kinds of objects in a set of pictures (all images that feature a specific logo), detect facial attributes such as whether people are smiling or their eyes are open or shut, and sift through hundreds of images to make a collage of the best photos, cropped to fit within a page. Moreover, they work in real time, opening up new avenues for improvement in devices like smart home assistants that are constantly aware of their surroundings.

Face grouping: Accurately recognize and group individuals.

Face grouping: Accurately recognize and group individuals.

 Some of these capabilities are unique to HP. For example, by installing HP’s Pixel Intelligence software, a print service provider could analyze a large collection of images such as wedding photos in a batch operation, automatically group faces of the same person together with high accuracy, and select the photos with the best facial image quality for inclusion in a wedding photobook. Other capabilities are offered by competing companies, but only through their own cloud servers.

“With Pixel Intelligence, you can host the processing software within your own data center, saving you time and bandwidth while keeping your data secure and private,” notes Dr. Lin.

The Pixel Intelligence portfolio is the result of close collaboration with HP’s Print Software Platform team. It was launched on HP’s developer site last fall and was presented to the Digital Solutions Cooperative (DSCOOP) of HP technology users in Phoenix this spring, with a repeat presentation at DSCOOP in Lyon in early June. It’s already drawing interest from photo services companies interested in increasing personalized print workflows and automating the creation of high quality print products.

“A lot of these companies don’t have the resources for artificial intelligence research, so they are very interested in licensing our technology,” Dr. Lin says. “We’re also hearing from companies in a wide range of industries - retail, health, security, and finance, for example – that want to know more about these capabilities.” 

Computer vision remains a continuing area of interest in HP Labs. Dr. Lin and her colleagues keep improving their existing algorithms and will add any new algorithms to the Pixel Intelligence portfolio.

They are also tackling new challenges, such as improving techniques for identifying 3D objects and applying advanced computer vision to ambient computing technologies that anticipate human needs and proactively address them.

“We are already placing image-gathering sensors in more devices and more places than ever before,” observes Dr. Lin. “So we see a lot of potential for this technology in future smart home, smart office, and mobile applications.”