Name
Abstract | ||
|---|---|---|
Temporal action detection in untrimmed video is an important and challenging task in computer vision. In this paper, a straightforward and efficient regression model is proposed by us to detect action instance and refine action interval in long untrimmed videos. We train a single 3D Convolutional Networks (3D ConvNets) jointly with two sibling output layers: a classification layer to predict the class label and a temporal interval regression layer to modify the temporal localization of input proposal. We also introduce an effective method to sample negative and positive proposals which are discriminative to feature extractor and classifier during training. On THUMOS 2014 dataset, our method achieves competitive performance compared with recent state-of-the-art methods. |
Year | DOI | Venue |
|---|---|---|
2017 | 10.1007/978-3-319-77380-3_25 | ADVANCES IN MULTIMEDIA INFORMATION PROCESSING - PCM 2017, PT I |
Keywords | Field | DocType |
Action detection,Temporal interval regression,3D ConvNets,Multi-task loss | Regression,Temporal interval,Pattern recognition,Regression analysis,Computer science,Effective method,Extractor,Artificial intelligence,Classifier (linguistics),Discriminative model | Conference |
Volume | ISSN | Citations |
10735 | 0302-9743 | 0 |
PageRank | References | Authors |
0.34 | 9 | 4 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Qing Wang | 1 | 345 | 76.64 |
| Laiyun Qing | 2 | 337 | 24.66 |
| Jun Miao | 3 | 11 | 3.31 |
| Lijuan Duan | 4 | 215 | 26.13 |