Computer Vision, Living on the Edge
We capture, organize and enjoy our photographs and videos on our personal devices, a.k.a. the edge. Mobius Labs licenses out lightweight, state-of-the-art computer vision algorithms that empower the edge with true visual intelligence.
Our Solutions
Mobius Vision SDK
Available as on device SDK or on premise installation
On Device SDK
Our lightweight SDK is engineered to run on mobile devices with fast processing (20 ms per image on Snapdragon 820 chipsets), small size and a low energy footprint.
On Premise Installation
We license products for processing vast amounts of images directly on your local machines at your premise. This enables smooth integration and control over the data flow.
Customized Training: Anyone can now train an AI model!
Everyone has different visual preferences — so you should be able to teach a machine what you like. Want to find the photo that best communicates your style to your audience? Or are you looking for your personal highlights summary of your last trip? With the customized training feature of Mobius Vision SDK, anyone can now train an AI model by selecting a few pictures they want to see, and a few they do not want to see. Your customized model is ready in just a few clicks!
1
2
3
4
5
6
7
8
9
//Train your custom model on positive/negative images //Run training in a background AsyncTask protected Void doInBackground(Void... VoidInput) { MobiusSDK.trainCustomModel(customModelKey, positiveImages, negativeImages, new AsyncTaskTrainingProgressCallback(this) ); }
Image and Video Keywording: Predict over 5000 keywords
Our SDK includes a keywording model that can recognize over 5000 objects, emotions and actions, which can be used to efficiently tag, organize and categorize large image and video collections, and enables users to quickly retrieve visual content.
1
2
3
//Predict keywords using the Mobius model from a Bitmap ArrayList<Keyword> keywords = MobiusSDK.predictKeywords(bitmap).getThresholdedResult();
Similarity Search: Match style and content
They say a picture is worth a thousand words. Mobius Vision is able to search for images that are similar both in style and content. Simply provide a reference image, and let our powerful algorithms do the magic.
1
2
3
4
5
6
//Get visually similar images from a Bitmap and a set of features int flag = MobiusSDK.SEARCH_KEYWORDING_ONLY; //Search only based on // keywording information ArrayList<Pair<?,Float>> result = MobiusSDK.getSortedSearchResults(Bitmap bitmap, Map<?,float[]> cachedFeatures, flag)
Aesthetics: Find the best photos hidden in your photo collections
Mobius Vision comes shipped with a pre-trained, powerful aesthetics model that can be used for a variety of tasks, including sorting large image collections, finding the best images, as well as filtering out bad content.
1
2
3
//Predict the aesthetics of an image from a Bitmap float aestheticsScore = MobiusSDK.predictAesthetics(bitmap).getScore();
Engineered for...
Speed
Our on-device models leverage GPU and DSP in mobile chipset hardware, and clock 18 millisecond to process an image/frame on a decent smartphone.
Privacy
All the computation happens on-device, and your data never leaves your device. This makes sure that your privacy is preserved — always.
Simplicity
A few lines of code and the system is up and running — no installation hassle anymore! Engineers without knowledge of computer vision can easily integrate our solutions.
Industries
Smart camera assistants
Our software lives on the edge to empower OEMs to offer exciting features to mobile phone and camera users. Recognizing key concepts in images to switch to a shooting mode or suggesting optimal cropping of visual data.
Empower app users
Use our custom models to build personalized visual experiences for your users on their local devices. Mobius Vision can sift through large collections of images and videos in seconds, uncovering visual content that suits your app’s style and intent.
Organize your image data
The ability to serve the right content to a customer is key in any stock photo agency. Our powerful aesthetics model can be used to sort large bodies of visual data. Is your customer after a specific style? Then you can retrain the model to suit their style.