Study Toolkit for Human Perception of Videomaterial

Schuler/Hauser

Our project aims towards creating a modular solution for data set creation, online study creation, and statistical and exploratory data analysis. We call this our toolkit. The focus lies on video material, where a human rater should rate different excerpts based on their subjective perception.

The first part of this three-step project contains the assisted creation of data sets. We want to enable an easy conversion from raw data to data sets that contain modifications of different attributes. Examples of different kinds of modifications are editing single frames of a video, changing the speed of spoken/visual parts, or simply applying different encodings and decodings. Our goal for this part is to create a data set for ourselves with a vast amount of different modifications for dubbing research. This was the initial idea that led to our project.

The second part is the creation of an online study. The data for this study either comes from the first part or another source. The modularity of the toolkit enables to use each part separately, given the correct input formats. The online study will be either local or online and will offer a HTML page with integrated media data and scales for rating the excerpts based on their modified attributes.

The third part is the analysis of results generated in part two or from another source. Explanations and examples will guide a user toward statistical research. To get a feeling for the results and explore what and how the participants rated, exploratory graphs could be auto-generated. In addition, there will be the possibility for basic statistical analysis such as detecting outliers be it on the side of the tested data or the side of a participant rating drastically different than the rest. Analysis for this part is mostly done in the form of graphs but can be of other modalities, such as numerical if fitting. 

Creating such a toolkit solution is not trivial. Especially when there is a lot of established software that is able to do the mentioned parts individually. We strive to combine many different solutions and advance them to get a comprehensible toolkit with a modular structure enabling an easy entry and exit between the parts. Together with explanations and examples, students might see this to be an ideal entry point into the field of empirical research.

And here is the student group in their own words: