Please use this identifier to cite or link to this item:
https://scholarbank.nus.edu.sg/handle/10635/134421
Title: | FEATURE MINING IN TEMPORAL HUMAN ACTION LOCALIZATION | Authors: | YUAN JUN | Keywords: | Action, Temporal Localization | Issue Date: | 18-Aug-2016 | Citation: | YUAN JUN (2016-08-18). FEATURE MINING IN TEMPORAL HUMAN ACTION LOCALIZATION. ScholarBank@NUS Repository. | Abstract: | Temporal human action localization aims at detecting and labeling actions in untrimmed video sequences, which is a generalization of the traditional action classification problem and better models the real-life video content analysis. However, the localization problem is difficult due to the uncertainties in action occurrence and is still being developed at a preliminary stage. This research proposes the Pyramid of Score Distribution Features (PSDF) to address the uncertainties of action position, duration and label in untrimmed videos. The PSDF carries heuristic information of action occurrence, and can be calculated via an efficient recursive addition. The temporal contexts are further modeled by two recurrent structures, the Elman-Net and LSTM to further enhance the performance of action localization. | URI: | http://scholarbank.nus.edu.sg/handle/10635/134421 |
Appears in Collections: | Ph.D Theses (Open) |
Show full item record
Files in This Item:
File | Description | Size | Format | Access Settings | Version | |
---|---|---|---|---|---|---|
simu.avi | 8.27 MB | AVI | OPEN | None | Preview online | |
thesis.pdf | 3.58 MB | Adobe PDF | OPEN | None | View/Download |
Google ScholarTM
Check
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.