We are happy to share the new version of the MICO Platform (1.2 release). Along with several changes to the core services, the new version also features an extended set of extractors to build analysis pipelines.

The release is available from the download page, and we invite anyone interested to try out the various pipelines for content analysis.

In the current version, three different analysis pipelines are available:

To simplify switching between different pipelines, a new configuration service was developed. By choosing the current use-case, one can easily launch the required extractors for the analysis process.

In the backend some core components have been extended:

  • For persistence, we added a Hadoop HDFS storage module to allow distributed and fault-tolerant storage of the content items. HDFS further allows distributing the data for the analysis process over multiple servers to increase performance.
  • For meta-data storage, we added Anno4j – a simple and convenient API to read and write the W3C Web Annotation Data Model  and W3C Open Annotation Data Model.
  • For advanced querying we added sparql-mm, a multimedia extension for SPARQL 1.1 that allows temporal and spacial queries on multimedia data.


In the near future, we plan to add several additional extractors to the platform and extend the available pipelines to support all the use-cases from the MICO Project.

The extractors in the queue are:

  • Extended Animal detection: currently, the animal detection only tells you whether an image contains an animal or not. The plan for the extended version is that the extractor also detects the species of the animal (e.g. hooved animal).
  • Speech-To-Text: this extractor  transcribes the audio-line of a video into text, so that it can be further processed by text-analysis modules such as the Sentiment Analysis.
  • Sentiment Analysis extracts the general sentiment (or mood) of a text – whether the text is positive, negative or neutral.

We are currently also developing an improved broker. The current version only allows simple static pipelines that do not cover all requirements of the use cases. In order to support real cross-media analysis, the new broker will allow dynamic configuration of pipelines with complex routing.