Vision Engine
Polarr Vision Engine (U.S. Patent Pending) includes the algorithms used in Polarr’s own applications to enable licensees to build unique experience for their customers. The engine is under active development, for latest features or other platforms, please contact us.
Android Development
Polarr Photo Editing SDK
Overview
Polarr Photo Editing SDK includes Polarr's proprietary auto-enhancement algorithms as well as all filter effects and individual adjustments from Polarr Photo Editor. It also includes face feature detection and facial feature adjustments as seen in Polarr's face editing tools. The SDK runs offline and requires local GPU to render images.
Use cases
- Automatically enhance photos
- Automatically recognize and enhance faces within photos (ex. skin smoothing, teeth whitening, etc)
- Globally edit photos for parameters like color temperature, blur, lighting, HSL, denoise, dehaze, etc.
- Locally edit photos via brush mask, color mask, radial mask or gradient mask
- Remove objects within photos via magic eraser
- Share and import custom filters
Customization
- Custom art brush effects
- Custom filter effects
- Rendering speed optimization for target frame rate per second
Performance & Requirements
- Android version 14 (4.0.3) and above
- OpenGL ES 3.0 and above
Polarr Album SDK
Overview
Polarr Album SDK includes algorithms for photo aesthetics ranking, similarity grouping, object detection, duplicate removal, face grouping and identification, as well as auto enhancement. The SDK can operate on high volume of image inputs, and replies on machine learning models developed by Polarr and included in the SDK. Compared to cloud based image APIs, Polarr Album SDK operates completely offline without requirement to access internet, and utilizes local CPU and GPU (when available) for all computations - so that you never need to worry about data transaction between your company and Polarr.
Use cases
- Cluster of similar photos based on deep features
- Rank photos based on aesthetics such as composition, clarity, exposure, human emotions, etc
- Classify images by producing one tag for the most dominant object in the image
- Detect objects within an image and produce relevant tags and bounding boxes
- Recognize faces and cluster photos/events by faces
- Caption of individual photos or groups of photos to articulate what what’s happening in the photos
- Construct relationship between events, people, locations given a set of photos"
Customization
- Recognize and detect specific items
- Tune algorithms for specific best photo selection
- Computing speed optimization for amount of photos processed per second
Performance & Requirements
- Android version 14 (4.0.3) and above
- No GPU required, multiple DL, ML, and unsupervised learning models ready to run directly on CPU
Demo and Link
Polarr Camera SDK
Overview
Polarr Camera SDK is intended for providing real time machine guided composition，object detection, aesthetic rating, and rendering for complex filter effects as seen in Polarr Photo Editor.
Use cases
- Create interactive user experience for guided photo composition
- Application use cases for real-time object detection
- Rate video stream in terms of quality, aesthetic, lighting, composition in realtime
- Render 20MP image in over 5 frames per second using prebuilt filters in C++
- Render 1080p images in over 60 frames per second using prebuilt filters in Java
Customization
- Customizable classes of detectable objects
- Customizable control for machine guided composition
- Custom filter effects
- Rendering speed optimization for target frame rate per second
Performance & Requirements
- Android version 14 (4.0.3) and above
- OpenGL ES 2.0 support or above
Demo and Link
Contact us
Pricing
Testing and personal use
You can use the SDK from the Github links for free for personal use. However, these SDKs will expire automatically and won's install in production envionrments.
Commercial use
For commercial licensing, please reach out to info@polarr.co for pricing and case studies.
Polarr, Inc. All Rights Reserved 2018.
