M1 Master 360° Director's Cut Analyzing visual attention in 360° Storytelling

Team

  • Anton Schöps
  • Sebastian Pütz

Supervision

Prof. Dr. Sebastian Knorr

Development Process

Our project followed an agile development approach. We utilized GitHub Projects to implement a Kanban board, which allowed us to organize our work into, manageable tickets. Each ticket represented a specific task or feature. The Board consisted of the stages “To Do”, “In Progress” to “Done,” providing us with a good overview of the projects progress. Additionally, we also made a “Nice To Have” stage for potential new features.

Initially, we created a comprehensive roadmap that outlined critical milestones for each week. However, as development progressed, we encountered challenges that caused us to fall behind our timeline. The roadmap quickly became outdated as tasks took longer than anticipated, forcing us to adapt our approach. In order to meet our final delivery date, we had to prioritize essential features and work under strict deadlines, which consequently led to a more intense development phase toward the end of the project.

While our initial planning provided a good foundation, this experience highlighted the importance of building buffer time into project schedules and the reality that development timelines often require flexibility and adjustment.

App Features

Desktop GUI

We created a the graphical inteface with QTs by using PySide. It features a sidebar, video controlls and different views to analyze the saliency.

Views Options

In the app you have the options between different views. Under the videos you can view a Map that represent if the view direction mateches the directors cut or view the saliency map for the current frame. It is also possible to display the saliency and the directors cut as an overlay on the video.

Predictors

Choose between different predictors algorithms: Choose between fast or slower but more accurates algorithms.

Highlight Brush

Use the spotlight brush to brigthen a spot in the video and see the effects on re-predictions.