|
|
Video Feature for Classification
Behúň, Kamil ; Herout, Adam (referee) ; Hradiš, Michal (advisor)
This thesis compares hand-designed features with features learned by feature learning methods in video classification. The features learned by Principal Component Analysis whitening, Independent subspace analysis and Sparse Autoencoders were tested in a standard Bag of Visual Word classification paradigm replacing hand-designed features (e.g. SIFT, HOG, HOF). The classification performance was measured on Human Motion DataBase and YouTube Action Data Set. Learned features showed better performance than the hand-desined features. The combination of hand-designed features and learned features by Multiple Kernel Learning method showed even better performance, including cases when hand-designed features and learned features achieved not so good performance separately.
|
|
|
Tool for Collection and Processing of Sports Videos
Bahník, Marek ; Zemčík, Pavel (referee) ; Herout, Adam (advisor)
The aim of this thesis is to create a tool in Python for downloading videos from YouTube and parsing them into usable form for possible neural network training. Videos are split into separate clips using proprietary method of comparing two frames around suspect frame. These clips are furthermore classified on a~scale of relevance using optical flow. To download quality videos from YouTube, user has to input proper search word and custom filters. This tool is not fully automated and optimized, but it is designet to make dataset creation much faster and simpler. In general user has to intervene manually and check output data.
|
|
|
Video Feature for Classification
Behúň, Kamil ; Herout, Adam (referee) ; Hradiš, Michal (advisor)
This thesis compares hand-designed features with features learned by feature learning methods in video classification. The features learned by Principal Component Analysis whitening, Independent subspace analysis and Sparse Autoencoders were tested in a standard Bag of Visual Word classification paradigm replacing hand-designed features (e.g. SIFT, HOG, HOF). The classification performance was measured on Human Motion DataBase and YouTube Action Data Set. Learned features showed better performance than the hand-desined features. The combination of hand-designed features and learned features by Multiple Kernel Learning method showed even better performance, including cases when hand-designed features and learned features achieved not so good performance separately.
|
guest ::
