AdA Annotation Explorer v1.1 (Nov 2021)

[Annotations online: ]
Legend Colors Scale Charts

Welcome to the AdA Annotation Explorer,

the current prototype of our web-based application for querying, analyzing and visualizing semantic video annotations.
The Annotation Explorer is a joint work of the Semantic Technologies Group of HPI Potsdam, the film studies department
of FU Berlin, and Joscha Jäger the author of FrameTrail. Please visit our AdA Project website for more information.
To visualize annotations, you can either query one or more annotation types from one or more movies and/or scenes, or search for text and predefined values.
Be careful when selecting all scenes or all types, as a large number of annotations will slow down your browser.
Please note that for copyright reasons we can only grant public access to the public domain video "Occupy Wall Street".
The application does not work with Internet Explorer.

If you have any technical questions, please contact Henning Agt-Rickauer. If you have any questions about the annotations or the corpus, please contact the FU Berlin team.

Click to load an example query:

AdA-Filmontology & Video Player Interaction – ‘Occupy Wall Street’

Full range of manual annotations as well as a selection of automatically generated annotations of “Occupy Wall Street”, an activist video that was released with creative commons and can be accessed unrestrictedly. This player view allows to explore the broad range of AdA-Filmontology as well as the different modes of visualisation and annotation-player interaction in the Annotation Explorer.

Detailed Video Comparison – Two Occupy-Wall-Street Videos

Comparative view of two accessible video files that offers unregistered users further insight into the functionality of the Annotation Explorer using two video players at once.

Combined Value Search – Sad Music + Close Field Sizes

Example for the ‘Value Search’ that allows to find recurring patterns of synchronous annotations across a corpus of different films or within a specific film. Here: music that was perceived as ‘sad’ in combination with close field sizes.

Full-Text Search – ‘Wall Street’

Full text search for the term “wall street” in list view. Manual and automatically generated annotations that mention wall street verbally or in which wall street was identified as setting or topic.

Full-Text Search – 'Ship' (esp. Automated Image Captioning Results)

Player view of a full text search for “ship”. Among other search results of the automated image captioning (im2txt) that shows various shots of ships within the corpus. Such a search is for example a useful starting point for researching the metaphoric and metonymic dimension of films on the financial crisis. Videos are only partially accessible for external users due to copyrights.

Macro Dynamics Full-Length Feature Film – ‘The Company Men’

Macro view of annotation types that allows to analyze and compare various dynamics on the level of whole films. Shot length, optical flow, use of music, image intrinsic movement and other dimensions are juxtaposed.

Automatically Generated Annotations – ‘The Margin Call’

Overview of all automatically generated annotations for a full feature film. Automatic speech recognition, concept detection (densecap, neuraltalk2 & im2txt), optical flow analysis, shot detection, waveform of the volume (soundenvelope) and sound classification (yamnet).

Comparison of Automatically Generated Annotations for several Feature Films

Comparative view of automatically generated annotations for a selection of different films. Optical flow analysis, shot detection, waveform of the volume (soundenvelope) and sound classification (yamnet).

Automatically Generated vs Manual Annotations – Segments of Music and Speech

Comparative view of automatically generated annotations and independently manufactured manual annotations that allow to explore differences and similarities in human and machine perception.

 Image Search Results
Total Items: 0