| Popis: |
Event searching in videos is important task for video genre classification, event detection, content-based video searching, etc. Being able to search for a particular event without going through the whole video saves lots of time for the users. It also helps in indexing the videos based on an event in it, so that users can retrieve videos with same event efficiently. The proposed approach is three-step approach. The approach uses the map-reduce concept to decrease the overall processing time for a video using deep learning. As the number of frames in a video can be large but events are limited so to solve this issue, we have used Eratosthenes sieve based key-frame extraction technique for extracting key-frames that describe the video efficiently, then from the key-frames, we have extracted the events in them, and then indexing of the events is done after finding event boundaries. |
