Conference proceeding
W-TALC: Weakly-Supervised Temporal Activity Localization and Classification
COMPUTER VISION - ECCV 2018, PT IV, Vol.11208, pp.588-607
Lecture Notes in Computer Science
01/01/2018
DOI: 10.1007/978-3-030-01225-0_35
Abstract
Most activity localization methods in the literature suffer from the burden of frame-wise annotation requirement. Learning from weak labels may be a potential solution towards reducing such manual labeling effort. Recent years have witnessed a substantial influx of tagged videos on the Internet, which can serve as a rich source of weakly-supervised training data. Specifically, the correlations between videos with similar tags can be utilized to temporally localize the activities. Towards this goal, we present W-TALC, a Weakly-supervised Temporal Activity Localization and Classification framework using only video-level labels. The proposed network can be divided into two sub-networks, namely the Two-Stream based feature extractor network and a weakly-supervised module, which we learn by optimizing two complimentary loss functions. Qualitative and quantitative results on two challenging datasets - Thumos14 and ActivityNet1.2, demonstrate that the proposed method is able to detect activities at a fine granularity and achieve better performance than current state-of-the-art methods.
Details
- Title: Subtitle
- W-TALC: Weakly-Supervised Temporal Activity Localization and Classification
- Creators
- Sujoy Paul - Univ Calif Riverside, Riverside, CA 92521 USASourya Roy - University of California SystemAmit K. Roy-Chowdhury - University of California System
- Contributors
- V Ferrari (Editor)M Hebert (Editor)C Sminchisescu (Editor)Y Weiss (Editor)
- Resource Type
- Conference proceeding
- Publication Details
- COMPUTER VISION - ECCV 2018, PT IV, Vol.11208, pp.588-607
- Publisher
- Springer Nature
- Series
- Lecture Notes in Computer Science
- DOI
- 10.1007/978-3-030-01225-0_35
- ISSN
- 0302-9743
- eISSN
- 1611-3349
- Number of pages
- 20
- Grant note
- IIS-1724341 / NSF; National Science Foundation (NSF) Mayachitra Inc N0001415-C5113 / ONR; Office of Naval Research
- Language
- English
- Date published
- 01/01/2018
- Academic Unit
- Computer Science
- Record Identifier
- 9984446556002771
Metrics
1 Record Views