Supervisor: Daniel Helm, Martin Kampel
Automatic video content analysis is needed in many different domains such as historical film preservation or film archiving. Due to the enormous quantity of available visual media data, automatic approaches shall be used in order to find specific content in videos or create new visual representations of specific domains.
A first and fundamental step for automatic video content analysis is shot boundary detection. This mechanism aims to split a professional or semi-professional created video in its basic components: the so-called shots. One video consists of several shots which are concatenated with shot boundaries. There exist two main types of boundaries: the abrupt transitions (AT) and the gradual transitions (GT) such as wipes, dissolves or fades. In order to detect shot boundaries and finally to split a video in its basic shots, automatic processes such as traditional computer vision algorithms as well as deep learning-based approaches shall be applied.
The goal of this practical work is as follows:
For establishing an automatic video analysis platform, a “Shot Detection module”, shall be developed. This module shall be able to detect abrupt transitions as well as gradual transitions in given input videos in order to split videos in their basic shots.
- Literature Review – getting to know the algorithms
- Data Preparation
- Implementation and optimization of state-of-the-art approaches
- Creation of a deployable software package
- Evaluation of implemented algorithms
- Written Report/Thesis and final presentation
- Basic knowledge in computer vision
- Experience in Python, C++ and Matlab
- Interest in Deep Learning (Tensorflow, Keras, PyTorch) and Machine Learning