What can I do with Tator?
At its core, Tator is a framework for end to end video and image analytics. It takes care of a lot of things that make working with video difficult so you can focus on the thing that matters - turning videos into information - in a fully customizable way. There are a lot of pieces to Tator that make this possible, and therefore a lot of individual capabilities that can help your organization succeed with video.
Digital asset management
Video, images, audio, and generic files can be uploaded and hosted on Tator, making all your digital assets accessible to your team. Media can be organized via projects, folders, and searchable custom attributes. For a quick look, media can be viewed with static and animated thumbnails, or media can be played at multiple resolutions and rates directly in the platform. Tator handles synchronized videos such as stereo or multi-camera monitoring with multi-stream videos and playback. Tator supports individual and bulk file downloads, making it easy to export data to other native or cloud based tools. Tator has a customizable, scalable video encoding capability that allows definition of optimized downloadable and streaming media formats, ensuring the right balance between storage and quality for your use case. Customized media lifecycle strategies ensure you are never spending more on storage than necessary. Read more...
Data management
Tator has a flexible metadata system that can describe virtually any concept in video or imagery, allowing you to build the raw metadata needed to conduct a useful analysis. Through its system of users, layers, versions, and change logs, Tator provides the tools needed to know the provenance of your data - every piece of data created, edited, or deleted is tracked, providing full traceability down to the level of individual attributes. Database performance is tuned for large queries, allowing the retrieval of tens of thousands of annotations in seconds. All metadata is indexed in Elasticsearch, allowing for fast aggregation operations, search, and retrieval. Complex labeling tasks are facilitated with hierarchical label tree definitions. Read more...
Extract, transform, load
Tator facilitates video ETL workflows through its various media extraction utilities. Through its REST API, you can quickly extract frame or localization imagery at a precise frame number, enabling workflows such as computer vision algorithm training or dataset exploration. Tator also supports export of animated sequences, including extraction of video clips and animated sequences of track localizations. Read more...
Video and image annotation
Through a flexible user interface, your team can draw dots, lines, boxes, and polygons, label frame ranges and discrete events, and create tracks; each of these can be described in place upon creation with a customizable set of attributes. Through custom front-end applets, annotation can be accelerated with AI assistance using standing inference servers or client side processing. Quality assurance and quality control workflows are facilitated through annotation layers and permission systems. Frame accurate playback and seek, frame stepping, and frame scrubbing features in combination with real time annotation drawing allow reviewers to do their job as efficiently as possible. Multi-stream videos allow annotation of multiple synchronized videos at once. Convenience features such as bookmarks, favorites, and link sharing help annotators stay on top of their workloads. Read more...
Simplified MLOps
Tator enables scalable batch inference of AI-based computer vision algorithms with one click as well as in-line inference on video frames or images through the annotation view. Through its use of Argo Workflows, not only computer vision but virtually any asynchronous work can be executed with Tator, including tasks such as report generation. Workloads can be tracked through the user interface, or users can be notified upon workload completion via email. Workloads can be executed on other services, such as AWS Lambda or AWS SageMaker, by delegating from a lightweight Argo workflow. Tator also allows organizations to register their own processing resources for remote execution of asynchronous workloads, giving the flexibility to fully leverage on-premise resources or cloud processing credits or discounts with different cloud providers. Read more...
Visual analytics
Tator facilitates visualization and analysis through a combination of built-in analytics tools and a custom dashboard capability. With Tator's localization analytics view, reviewers can quickly search, review, and correct localization metadata. A dedicated bulk corrections interface allows users to quickly apply modifications at the media level. The summary dashboard allows users to visualize and export a tabulation of a filtered dataset within a project. Through custom dashboards, interactive applets can be registered with a project that analyze raw metadata and display results. They allow use of third party visualization libraries such as D3 as well as import of cross-origin data so they can work with other tools and data sources. Read more...