08-09-2016, 10:50 AM
1453747826-abstract.pdf (Size: 318.76 KB / Downloads: 5)
Abstract:
Given a set of text files and media files, the project aims at finding an association among them
and to cluster them on the basis of discovered association. The approach is to generate new
metadata and identify the features using machine learning techniques and then use a similarity
technique to cluster the files. The term 'multidimensional' is used to signify the multiple features
used for clustering.
The project involves 5 basic modules as
1. Metadata analysis: - Extracting relevant metadata.
2. Image processing: - Identifying discrete objects in an image (using neural networks) as new
metadata.
3. Audio processing: - Finding and updating the missing metadata and lyrics depending on the
availability using online databases.
4. Video processing: - Segregation of frames and audio. Frames and audio are being processed
by their respective modules and their collective result is used to generate new metadata.
5. Clustering: - Clustering based on similarity of newly generated metadata as well as traditional
metadata.