MICO — Towards Contextual Media Analysis

15
3rd International Workshop on Linked Media, WWW 2015, Florence Sergio Fernández, Sebastian Schaffert, Thomas Kurz MICO — Towards Contextual Media Analysis

Transcript of MICO — Towards Contextual Media Analysis

Page 1: MICO — Towards Contextual Media Analysis

3rd International Workshop on Linked Media, WWW 2015, Florence

Sergio Fernández, Sebastian Schaffert, Thomas Kurz

MICO — Towards Contextual Media Analysis

Page 2: MICO — Towards Contextual Media Analysis

© MICO Project. Funded by the European Commission FP7 (grant agreement no: 610480). No reproduction without written permission.

The MICO project

Page 3: MICO — Towards Contextual Media Analysis

© MICO Project. Funded by the European Commission FP7 (grant agreement no: 610480). No reproduction without written permission.

The MICO project idea

Page 4: MICO — Towards Contextual Media Analysis

© MICO Project. Funded by the European Commission FP7 (grant agreement no: 610480). No reproduction without written permission.

The MICO project idea

Page 5: MICO — Towards Contextual Media Analysis

© MICO Project. Funded by the European Commission FP7 (grant agreement no: 610480). No reproduction without written permission.

The MICO project idea

● provide an architecture to analyze media in context by orchestrating various analysis components for several media types

● provide all necessary technologies for a distributed analysis workflow: extractors, orchestration, schema, querying

● contribute the core software results as components under business-friendly OSS licenses (building on projects such as Apache Marmotta), but allowing inclusion of both OSS and closed-source extractors to simplify the use of the technology in industrial products.

Page 6: MICO — Towards Contextual Media Analysis

© MICO Project. Funded by the European Commission FP7 (grant agreement no: 610480). No reproduction without written permission.

Requirements in short■ large content items (video, audio), located on

different storage systems■ composite content items, consisting of a collection

of related media objects (e.g. website with video and images)

■ independent analysis components, each contributing a small part to the overall analysis resulto Java and C/C++ components, as analysis services

are typically implemented in the language that best serves the job

o user feedback, where users might be involved in the analysis process as “crowd workers” still to be defined

■ metadata stored in RDF following a cross-media modelo external applications can further process the

metadatao querying of metadata, content and context in a

unified way using SPARQL

Page 7: MICO — Towards Contextual Media Analysis

© MICO Project. Funded by the European Commission FP7 (grant agreement no: 610480). No reproduction without written permission.

Architecture (High Level)

Page 8: MICO — Towards Contextual Media Analysis

© MICO Project. Funded by the European Commission FP7 (grant agreement no: 610480). No reproduction without written permission.

Components overview

Page 9: MICO — Towards Contextual Media Analysis

© MICO Project. Funded by the European Commission FP7 (grant agreement no: 610480). No reproduction without written permission.

Platform Prototype

● collection of UNIX services (mostly Java and C++ analysers, FTP server, RabbitMQ Marmotta, PostgreSQL, etc)

● analysis services can be either started on the same machine or connect remotely

● distribution as Maven artifacts, VirtualBox image and Debian Linux packages

● available as open source http://mico-project.bitbucket.org

Page 10: MICO — Towards Contextual Media Analysis

© MICO Project. Funded by the European Commission FP7 (grant agreement no: 610480). No reproduction without written permission.

Get MICO running

1.Setup Debian Jessie

2.Add MICO Repository

3. Install MICO Platform

$ sudo add-apt-repository 'deb http://apt.mico-project.eu/ mico main contrib non-free'$ wget -O- http://apt.mico-project.eu/apt-repo.key | sudo apt-key add -$ sudo apt-get update$ sudo apt-get install mico-platform

https://bitbucket.org/mico-project/platform#markdown-header-debian-repository

Page 11: MICO — Towards Contextual Media Analysis

© MICO Project. Funded by the European Commission FP7 (grant agreement no: 610480). No reproduction without written permission.

DEMO

Page 12: MICO — Towards Contextual Media Analysis

© MICO Project. Funded by the European Commission FP7 (grant agreement no: 610480). No reproduction without written permission.

Using the API: Javafinal EventManager eventManager = new EventManagerImpl("192.168.56.101");eventManager.init();

final PersistenceService persistenceService = eventManager.getPersistenceService();final ContentItem ci = persistenceService.createContentItem();final Content cp = ci.createContentPart();cp.setType("text/plain");

IOUtils.copyLarge(new FileInputStream("example.txt"), cp.getOutputStream());

eventManager.injectContentItem(ci);

Page 13: MICO — Towards Contextual Media Analysis

© MICO Project. Funded by the European Commission FP7 (grant agreement no: 610480). No reproduction without written permission.

Using the API: C++EventManager eventManager(mico_host, mico_user, mico_pass);ContentItem* item = eventManager.getPersistenceService().createContentItem();

Content* c = item->createContentPart();c->setType("text/plain");int fd = open("example.txt", O_RDONLY);size_t len = st.st_size;char* buffer = (char*)mmap(NULL, len, PROT_READ, MAP_SHARED, fd, 0);std::ostream* os = c->getOutputStream();os->write(buffer, len);

eventManager.injectContentItem(*item);

https://bitbucket.org/mico-project/platform/src/master/api/c++/tools/mico_inject.cpp

Page 14: MICO — Towards Contextual Media Analysis

© MICO Project. Funded by the European Commission FP7 (grant agreement no: 610480). No reproduction without written permission.

Future Plans

● Refine messaging API

● Model API based on Open Annotation

Ontology

● SPARQL-MM extensions

● Specialized Data APIs

● and Stabilize

Next Release: May 22 2015

Page 15: MICO — Towards Contextual Media Analysis

© MICO Project. Funded by the European Commission FP7 (grant agreement no: 610480). No reproduction without written permission.

Thanks for your attention!

Sergio Fernández, Sebastian Schaffert, Thomas Kurz

[email protected]

Knowledge and Media TechnologiesSalzburg Researchhttp://www.salzburgresearch.at