what is Tensorgrid?
Processing media through state-of-the-art deep learning
TensorGrid rapidly processes your media using state-of-the-art deep learning technology to extract high level content and relationships. Relevant content can be found using the power of visual search, face recognition, custom classifiers, and Cubica's powerful visual recommender technology.
The system applies image analytics algorithms to tag media with rich, high level information such as scenes, objects, concepts, and identities. Its media analytics tools summarise content in simple, intuitive and interactive ways and the visual recommender helps users find content that may have been overlooked. Bespoke classifiers can also be trained in a few simple clicks using private data.
The platform scales to millions of images with ease by utilising both vertical (multi-CPU, multi-GPU) and horizontal (cluster) scaling. It can scale seamlessly to process high volumes of data.
TensorGrid can be deployed in a range of ways; in the cloud, on a private network, or on a standalone machine. Pre-configured multi-GPU TensorGrid servers are also available to allow users to hit the ground running.
Manual media analysis is a laborious, time consuming task. For investigative work, TensorGrid drastically reduces the time needed to review a collection of media.
It’s easy to overlook content when under time pressure. TensorGrid can help users truly understand and exploit media to its fullest potential. Searching for content becomes easy and fast, and the right media comes to the top of the stack.
Scalable backend processing
Web-based graphical interface
Visual filtering, sorting, and search
EXIF metadata, abstract object and scene tagging
Similar image search
Access to core functions via REST API
Analytics and charting
The visual recommender plug-in helps locate relevant content that might otherwise be overlooked. Simply give the recommender one or more examples of what you're looking for and it will create a dynamic classifier on the fly and recommend media to view.
Local content search
TensorGrid’s Local Search plug-in brings a step change in performance for finding similar content. Simply select a region of an image and the system will identify other similar regions from the entered dataset.
TensorGrid's deep learning-based face recognition plug-in identifies faces and identities in submitted media. Users can search for specific numbers of faces, specific people, and identify key identities and relationships (who knows who) in media with human level performance.
Text extraction and OCR
Extract key text from images, diagrams and drawings and perform multi-language deep learning-based optical character recognition (OCR), making the text within images identifiable and searchable.
Not safe for work (NSFW)
The NSFW plug-in provides the capability to detect and/or censor adult content. Users can rapidly scan media for content of concern and exercise duty of care by hiding or flagging the sensitive content.
Train your own high performance classifier to detect the content you are most interested in, and slot this classifier into your media ingest pipeline.
Talk to the experts
In addition to our off-the-shelf plug-ins, Cubica can help build tailored algorithms or plug-ins. Our team of pattern recognition experts can help with the entire process from concept to operational use. Talk to an expert today.