• LUNAR

    Media Monitoring Software
    • pic
      Detect logos in videos
    • pic
      Compute on a massive scale
    • pic
      Analyze Webmedia
    • pic
      Search millions of media files
    Search and analyze different media including video and audio - automate using a scalable solution! Learn more
  • LUNAR

    Image Recognition
    Logo Recognition in TV, webstreams and print-media
    • pic
      • Detect logos in videos
        (TV, webstreams, press/news and more)
      • High recognition rate
      • 1000x more efficient then human recognition
      • Detects logos on clothes, advertising boards, ...
      • Track logos in realtime

    Learn more
  • LUNAR

    Distributed Computing
    Distributed Computing with up to 10,000 nodes
    • pic
      • Supports up to 10,000 nodes
      • High availability
      • Monitoring Webfrontend
      • Integration using REST-API
      • Distributes all LUNAR software
    Learn more
  • LUNAR

    Webmedia Analysis
    Image and text analysis of webarticles with high data throughput
    • pic
      • Analyze web article texts and images
      • Use URLs from RSS-feeds
      • Supports content behind paywalls / age verification
      • High data throughput
      • Process further with LUNARs Image Recognition
    Learn more
  • LUNAR

    Media Indexing
    Media Indexing to compare millions of video-, image- and audiofiles
    • pic
      • Search within millions of media files
      • Supports different types of media
        (video, image, audio and text)
      • Compare images and video sequences
      • Efficient data reduction
    Learn more

LUNAR Media Monitoring

LUNAR is a modular, scalable solution ideal for Media Monitoring Applications which realizes processing and classification of many different media sources in a clustering environment.

pic

  • LUNAR Image Recognition and LUNAR Webmedia Analysis build a powerful system that generates valuable metadata information like the media exposure of sponsoring in video or webmedia. We constantly improve LUNAR and its components and create expansions and enhancements to keep our software solutions state-of-the-art.

    In addition to this we help integrate our software into the environments and workflows of our customers to realize their individual requirements.

LUNAR Image Recognition

International media-monitoring companies use our Image Recognition System to search for information in sponsoring within billions of images. Our software is robust, uses state-of-the-art detection algorithms and optimizes its recognition process using machine learning and statistical evaluation of thousands of data sets.
    • Finds logos in videos
    • High recognition rate
    • Tolerates scaling & rotation
    • Tolerates partially obstructed objects
    • Tracking with 30FPS in realtime
  • Detects and tracks logos in videos

    Our Image Recognition System is based on state-of-the-art routines to detect image information like a sponsoring logo in a real world video like a football match for example. Detected logos are tracked forwards and backwards in time using an efficient algorithm. We achieve recognition rates of up to 99% using our benchmarking material and we achieve similar rates in running production environments.
  • Detects and tracks logos in videos

    Our Image Recognition System is based on state-of-the-art routines to detect image information like a sponsoring logo in a real world video like a football match for example. Detected logos are tracked forwards and backwards in time using an efficient algorithm. We achieve recognition rates of up to 99% using our benchmarking material and we achieve similar rates in running production environments.
  • Scale- and rotation-invariant algorithms

    LUNAR detects logos which are actually a lot bigger or smaller within the video to be analyzed. We support differences in scale a lot larger than known techniques and also tolerate finding rotated logos with ±180°. This makes our software more robust against zooming and fast movements of the camera.
  • Scale- and rotation-invariant algorithms

    LUNAR detects logos which are actually a lot bigger or smaller within the video to be analyzed. We support differences in scale a lot larger than known techniques and also tolerate finding rotated logos with ±180°. This makes our software more robust against zooming and fast movements of the camera.
  • Tolerates obstructed and transformed objects

    Partially obstructed objects are found by our algorithms as well. A descriptive part of a logo can still be detected when only 20% of the total area of the logo is visible. Transformed image patches like a logo on clothes or advertising boards can be found by looking for different linear projections as well.
  • Tolerates obstructed and transformed objects

    Partially obstructed objects are found by our algorithms as well. A descriptive part of a logo can still be detected when only 20% of the total area of the logo is visible. Transformed image patches like a logo on clothes or advertising boards can be found by looking for different linear projections as well.
  • Fast analysis with full-frame precision

    Binärwerkstatt's Image Recognition System can not only compete in terms of quality but also speed: LUNAR can be used to analyze videostreams with 25 frames per second in realtime using only 1 PC! We calculate sequences of found logos with the timing-precision your process demands.
  • Fast analysis with full-frame precision

    Binärwerkstatt's Image Recognition System can not only compete in terms of quality but also speed: LUNAR can be used to analyze videostreams with 25 frames per second in realtime using only 1 PC! We are calculating sequences of found logos with the timing-precision your process demands.
  • Supports a large number of models

    LUNAR can distinguish between thousands of models/logos within an analysis of a video. Technically we support an unlimited number of models per analysis and run production clusters that analyze up to 2.500 models simultaneously.
  • Supports a large number of models

    LUNAR can distinguish between thousands of models within an analysis of a video. Technically we support an unlimited number of models per analysis and run production clusters that analyze up to 2.500 models simultaneously.
  • Supports custom media sources

    By supporting detecting information on a 1 FPS video or on single images it is possible to analyze other media like print-, web- or other media. We have gained expertise in analyzing different media sources.
  • Supports custom media sources

    By supporting detecting information on a 1 FPS video or on single images it is possible to analyze other media like print-, web- or other media. We have gained expertise in analyzing different media sources.

LUNAR Distributed Computing

Our Software Solution for Distributed Computing realizes a high daily throughput of video- and image-analyses. Large Systems consist of up to 10,000 Nodes joined together.

  • Utilizes hundreds of Nodes for your recognition tasks!

    We have been running our computer vision tasks on distributed systems for years now and have a great experience of developing and maintaining large systems consisting of hundreds of Nodes around the globe. Expanding clusters on demand and performing changes to the configuration is done without any downtime.

    High availability using Failover Mechanics and Master Redundancy

    We deliver very high availability with nearly zero downtime. We use many different failover mechanics to handle most known problems automatically. Rescheduling tasks because of a node hardware failure or electing a new master using Master Redundancy are just a few examples.

  • Utilizes hundreds of Nodes for your recognition tasks!

    We have been running our computer vision tasks on distributed systems for years now and have a great experience of developing and maintaining large systems consisting of hundreds of Nodes around the globe. Expanding clusters on demand and performing changes to the configuration is done without any downtime.
  • High availability using Failover Mechanics and Master Redundancy

    We deliver very high availability with nearly zero downtime. We use many different failover mechanics to handle most known problems automatically. Rescheduling tasks because of a node hardware failure or electing a new master using Master Redundancy are just a few examples.
  • pic
  • Easy-to-use RESTful-Webservice

    By using our JSON Restful-Webservice you can schedule thousands of analyses at a time, administer added analyses and fetch the results. Each analysis of a video can be prioritized in order to speed up selected analyses on demand. Statistics can be retrieved to see the daily throughput or the status of each currently running analysis. This enables you to use all the relevant information about the recognition process within your own management and datamining applications.
  • Easy-to-use RESTful-Webservice

    By using our JSON Restful-Webservice you can schedule thousands of analyses at a time, administer added analyses and fetch the results. Each analysis of a video can be prioritized in order to speed up selected analyses on demand. Statistics can be retrieved to see the daily throughput or the status of each currently running analysis. This enables you to use all the relevant information about the recognition process within your own management and datamining applications.
  • Monitoring Web-Frontend for quick-checking Health, Jobs and Nodes

    For quick-checking an operating Cluster you can use our Monitoring Web-Frontend. All you like to know can be retrieved by simply using your Webbrowser: You can see all nodes information like load, used memory, filesystem usage, and overall node health. You can see how many analyses have been processed the last 24 hours or the last week. Detailed information about an analysis like status, models, analysis time, video information and more are included as well.
  • Monitoring Web-Frontend for quick-checking Health, Jobs and Nodes

    For quick-checking an operating Cluster you can use our Monitoring Web-Frontend. All you like to know can be retrieved by simply using your Webbrowser: You can see all nodes information like load, used memory, filesystem usage, and overall node health. You can see how many analyses' have been processed the last 24 hours or the last week. Detailed information about an analysis like status, models, analysis time, video information and more are included as well.
  • Cluster Housing

    Either inhouse-clustering at your company's place, rented-clustering run and maintained on Binärwerkstatt's clusters or clusters running on common cloud-computing platforms via Linux Containers / Docker are supported.
  • Cluster Housing

    Either inhouse-clustering at your company's place, rented-clustering run and maintained on Binärwerkstatt's clusters or clusters running on common cloud-computing platforms via Linux Containers / Docker are supported.
  • Custom Tasks Support

    If your process demands custom tasks like transcoding for example we can support the distribution of these tasks within the system's architecture. We are expert in Computer Vision, Databases and Data-Mining and -Management and can help you in various fields like webscraping, video- and image-indexing or textmining and make it possible to compute on a massive scale.
  • Custom Tasks Support

    If your process demands custom tasks like transcoding for example we can support the distribution of these tasks within the system's architecture. We are expert in Computer Vision, Databases and Data-Mining and -Management and can help you in various fields like webscraping, video- and image-indexing or textmining and make it possible to compute on a massive scale.

LUNAR Webmedia Analysis

LUNAR's Web Analysis Applications make it possible to crawl large amounts of web-links and retrieve valuable metadata for further processing. With each crawl attempt we simulate browsing a web-link and distinguish between different contents and save it in speed- and size-optimized storage solutions. LUNAR's Webmedia Analysis Applications enable on-the-fly automation on classifying webmedia like detecting and classifying sponsoring logos or textual-content within web-articles.

  • Crawls hundreds of thousands of websites per day

    LUNAR's Webmedia Analysis Applications are ready for Distributed Computing and are used to aggregate media from the internet with huge daily throughput. Given a list of urls to important webmedia the system starts to retrieve, render and store the media to accelerated data structures like databases and distributed filesystems.

    Crawls websites by links from specialized RSS feeds

    Newsworthy web-articles can be retrieved by using RSS-feeds. It is possible to crawl thousands of web-articles efficiently that originate from just a few RSS-feeds. This way the system is able to retrieve huge amounts of topic sensitive webmedia with a minimum of required maintenance.

  • Crawls hundreds of thousands of websites per day

    LUNAR's Webmedia Analysis Applications are ready for Distributed Computing and are used to aggregate media from the internet with huge daily throughput. Given a list of urls to important webmedia the system starts to retrieve, render and store the media to accelerated data structures like databases and distributed filesystems.
  • Crawls websites by links from specialized RSS feeds

    Newsworthy web-articles can be retrieved by using RSS-feeds. It is possible to crawl thousands of web-articles efficiently that originate from just a few RSS-feeds. This way the system is able to retrieve huge amounts of topic sensitive webmedia with a minimum of required maintenance.
  • pic
  • Distinguishes between web-article elements

    LUNAR's Webmedia Analysis Applications classify web-articles on the fly by using known metrics and information density to gather information about what is relevant on a rendered webpage: It is able to distinguish between information like position, headline, textual-, image-, video- and audio-elements within or outside the article.
  • Distinguishes between web-article elements

    LUNAR's Webmedia Analysis Applications classify web-articles on the fly by using known metrics and information density to gather information about what is relevant on a rendered webpage: It is able to distinguish between information like position, headline, textual-, image-, video- and audio-elements within or outside the article.
  • Persistent cookie store

    Cookies from websites can be stored persistently and can be used in the crawling process. This way websites which pop up either an age- or account-verification (websites behind a paywall) can be processed without any problems.
  • Persistent cookie store

    Cookies from websites can be stored persistently and can be used in the crawling process. This way websites which pop up either an age- or account-verification (websites behind a paywall) can be processed without any problems.
  • Classification of webmedia

    LUNAR enables you to analyze crawled and stored webmedia with other classification systems like LUNAR's Image Recognition. Retrieved information like a matching sponsoring-logo for example can be linked to information retrieved from the article headline and other textual elements or the source from the RSS feed for example.
  • Classification of webmedia

    LUNAR enables you to analyze crawled and stored webmedia with other classification systems like LUNAR's Image Recognition. Retrieved information like a matching sponsoring-logo for example can be linked to information retrieved from the article headline and other textual elements or the source from the RSS feed for example.

LUNAR Media Indexing

LUNAR indexes text, video, image and audio-information to realize efficient, mixed searches on different datatypes. Results from LUNAR's Image Recognition or other external classification analyses can be included in searches as well by combining them using specialized metadata indexes.

  • Search within millions of images, videos and audiofiles

    LUNAR Media Indexing is able to search within millions of images and millions of videos. It is possible to find matching video sequences inside a video and locate the sequence, find audio sequences inside an audio-stream and locate its beginning and end.
  • Search within millions of images, videos and audiofiles

    LUNAR Media Indexing is able to search within millions of images and millions of videos. It is possible to find matching video sequences inside a video and locate the sequence, find audio sequences inside an audio-stream and locate its beginning and end.
  • Create indexes with various media types

    LUNAR supports the creation of indexes on videos, images, audio-streams and text. It is possible to combine searches on different media types using logical operators or to do custom searching within metadata retrieved by using speech recognition on audio or OCR on videos for example.
  • Create indexes with various media types

    LUNAR supports the creation of indexes on videos, images, audio-streams and text. It is possible to combine searches on different media types using logical operators or to do custom searching within metadata retrieved by using speech recognition on audio or OCR on videos for example.
  • pic
  • Eliminating duplicates

    Duplicates are mostly unwanted in the processing of data. Eliminating duplicates can reduce the workload extensively. LUNAR Media Indexing makes it possible to look for identical media and improves quality by selecting high resolution media for an OCR application for example and speeds up further processing by skipping duplicates - both for machine computing and humans.
  • Minimize workload by eliminating duplicates

    Duplicates are mostly unwanted in the processing of data. Eliminating duplicates can reduce the workload extensively. LUNAR Media Indexing makes it possible to look for identical media and improves quality by selecting high resolution media for an OCR application for example and speeds up further processing by skipping duplicates - both for machine computing and humans.
  • Efficiently solve classification problems

    LUNAR Media Indexing works well as a tool for solving classification problems. Our media indexes realize a fast reduction of data by using accelerated data structures. Further restricting media needed by the classification process is achieved by user-defined search-requests.
  • Efficiently solve classification problems

    LUNAR Media Indexing works well as a tool for solving classification problems. Our media indexes realize a fast reduction of data by using accelerated data structures. Further restricting media needed by the classification process is achieved by user-defined search-requests.