Atomica: Automatic movie classifier using audio-visual content analysis
Numerous studies with regards to classification of movies into genres are being conducted. Different techniques are used and implemented such as identifying the key components of the audio and visual aspects of the movie. Some of these techniques are the extraction of relevant information from the m...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | text |
| Language: | English |
| Published: |
Animo Repository
2010
|
| Subjects: | |
| Online Access: | https://animorepository.dlsu.edu.ph/etd_bachelors/11208 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | De La Salle University |
| Language: | English |
| Summary: | Numerous studies with regards to classification of movies into genres are being conducted. Different techniques are used and implemented such as identifying the key components of the audio and visual aspects of the movie. Some of these techniques are the extraction of relevant information from the movie which would help in classifying movies into genres.
In this research, an automated movie classification system was developed based on the audio-visual contents of the movie. The system uses audio and visual features to classify a certain movie according to its genre. Techniques such as frame extraction, luminance computation, visual disturbance and audio disturbance calculation, as well as shot segmentation are implemented for obtaining relevant data to be used in the classification of the movie if it is action, comedy or drama.
The features-based algorithm used in classifying action and non-action clips have 91.11% accuracy based on a test set consisting of 60 manually selected action movie clips and 120 manually selected non-action movie clips. In identifying comedy and drama genre, the feature-based algorithm used shows 94.17% accuracy on a set of clips composed of 60 manually selected comedy clips and 60 manually selected drama clips. The compilation of these algorithms shows 87.8% based on 180 manually selected clips composed of 60 action, 60 comedy and 60 drama clips. |
|---|