Polarr raises $11.5 million for offline, on-device computational photography

The Polarr team, here pictured at an offsite in Yosemite.

Image Credit: Polarr

Polarr, a four-year-old San Jose computer vision startup cofounded by Stanford graduate and Google veterans Borui Wang and Derek Yan, today announced that it has secured $11.5 million in series A funding led by Threshold Ventures, with participation from Cota Capital and Pear Ventures. Wang said the fresh capital — which brings its total raised to $13.5 million, according to Crunchbase — will be used to accelerate research and development; expand platform and service support; and grow its technology partnerships in drone, home appliance, ecommerce, and image storage verticals.

“As deep learning compute shifts from the cloud to edge devices, there is a growing opportunity to provide sophisticated and creative edge AI technologies to mobile devices,” said Wang, who serves as CEO. “This new round of financing is a tangible endorsement of our approach to enable and inspire everyone to make beautiful creations.”

Threshold Ventures’ Chris Kelley and Pear Ventures’ Mar Hershenson will join Polarr’s board of directors as part of the round. “Polarr’s expertise across design, hardware, and deep learning is really unique,” said Kelley. “Not only do they give consumers a way to leverage skills from the world’s best photographers, but pros can also use their tools to create something really special.”

Polarr’s bread and butter is its eponymous Polarr Vision Engine, a hardware-agnostic AI stack purpose-built to enable computational photography on a range of devices. It comprises a set of self-trained neural network models, each individually compressed and optimized for on-device storage, RAM usage, and power consumption constraints.

The Polarr Vision Engine underlies tech from Qualcomm, Oppo, Hover Camera, and others and powers Composition Guide, a real-time feature for the Samsung Galaxy S10’s native camera app that susses out optimal compositions and provides interactive guiding prompts. Moreover, it’s at the heart of Polarr’s trio of cross-platform photo editing apps for macOS, iOS, Android, and Windows 10: Polarr Photo Editor, Album+, and Deep Crop.

Photo Editor boasts a robust set of overlay and blending modes, in addition to dual lens effects and depth adjustment, intelligent automatic exposure and white balance adjustment, and masking — all contained in a customizable workspace with rearrangable icons. There are local adjustment options, like a border tool that auto-suggests colors based on the content of photos, plus features like face editing and swappable filters.

Polar says that Photo Editor has 4 million monthly users globally and over 20 million users in total.

Album+ offers a more passive experience — it automatically classifies and ranks photos of people, places, documents, objects, and receipts, and it can batch edit and delete hundreds of pictures or remove those that were poorly shot. As for Deep Crop, it leverages machine learning algorithms trained on data from “millions” of photographers around the world to suggest new photo angles.

Deep Crop is available for free, but Photo Editor subscriptions start at $2.49 per month or $23.99 per year, while Album+ starts at $1.99 per month or $12.99 per year.

Polarr’s software development kits (SDKs) come in three flavors: Photo Editing SDK, Polarr Album SDK, and Polarr Camera SDK. The Photo Editing SDK — which runs offline and requires a local GPU to render images — packs in auto-enhancing algorithms (e.g., object removal, denoising, skin smoothing, and teeth whitening), filter effects, and facial feature adjustments. Polarr Album SDK — which similarly runs offline and which taps local processor and graphics cores where available — includes AI models for photo aesthetics ranking, similarity grouping, object detection, face grouping and identification, and more. And the Polarr Camera SDK provides object detection, aesthetic ranking, and rendering for complex filter effects. (Polarr says it can render a 20-megapixel image in over 5 frames per second with prebuilt C++ filters, or 1080p images in over 60 frames per second using Java filters.)

All three SDKs are freely available for personal use from GitHub but expire automatically and don’t install in production environments.

Polarr’s core team of about 24 members is spread across offices in Shenzhen and Silicon Valley and includes graduates from educational institutions like Stanford, Carnegie Mellon, and Duke and former employees of companies like Microsoft, Google, Qualcomm, and Baidu.

Sign up for Funding Daily: Get the latest news in your inbox every weekday.