• Chinese Journal of Lasers
  • Vol. 46, Issue 11, 1109001 (2019)
Yunwen Huang1, Fei Wang2、*, Jinghong Li1, and Guorui Wang2
Author Affiliations
  • 1College of Information Science and Engineering, Northeastern University, Shenyang, Liaoning 110819, China
  • 2Faculty of Robot Science and Engineering, Northeastern University, Shenyang, Liaoning 110169, China
  • show less
    DOI: 10.3788/CJL201946.1109001 Cite this Article Set citation alerts
    Yunwen Huang, Fei Wang, Jinghong Li, Guorui Wang. Algorithm for Video Temporal Action Proposal Combining Watershed and Regression Networks[J]. Chinese Journal of Lasers, 2019, 46(11): 1109001 Copy Citation Text show less
    Whole architecture of temporal detection algorithm
    Fig. 1. Whole architecture of temporal detection algorithm
    Principle of improved watershed proposal algorithm
    Fig. 2. Principle of improved watershed proposal algorithm
    Structure of temporal pyramid of internal and extended regions
    Fig. 3. Structure of temporal pyramid of internal and extended regions
    Performances of candidate regions on Thumos 14. (a) AR-AN; (b) Recall-AN-tIoU
    Fig. 4. Performances of candidate regions on Thumos 14. (a) AR-AN; (b) Recall-AN-tIoU
    Schematic of candidate regions generated in each stage of two-stage regional proposal algorithm
    Fig. 5. Schematic of candidate regions generated in each stage of two-stage regional proposal algorithm
    Ablation study oftemporal context module, temporal pyramid module, and C3D unit-level feature
    Fig. 6. Ablation study oftemporal context module, temporal pyramid module, and C3D unit-level feature
    MethodDAPs+SVMSVMSCNNlocalizer
    DAPs[14]13.99.516.3
    Sparse-prop[15]7.88.115.3
    SST[16]15.923.0
    BSN[17]20.729.4
    SCNN-prop[10]7.614.019.0
    Watershed4.96.715.2
    Reg8.49.918.6
    Watershed+Reg24.723.837.2
    Table 1. Performances of different temporal proposal methods in subsequent localization in candidate regions on Thumos 14 %
    tIoU0.50.750.95Average
    Method in Ref. [3]42.283.760.0514.85
    BSN[17]46.4529.968.0230.03
    CDC[18]45.30260.2023.80
    TCN[19]23.58
    SCC[20]4017.904.7021.70
    Ours48.5831.748.7131.23
    Table 2. mAP of temporal localization coordinate of each method with different tIoU on ActivityNet v1.3 dataset %
    MethodAR-AN of 1000FPS
    DAPs[4]57.64134.30
    Sparse-prop[2]56.6010.20
    SST[3]60.27308
    CDC[18]500
    Proposal-1666.27423.15
    Proposal-3262.35760.84
    Proposal-w/o unit60.41129.40
    Table 3. Comparison of FPS and recall rate of different methods on Thumos 14 dataset
    Yunwen Huang, Fei Wang, Jinghong Li, Guorui Wang. Algorithm for Video Temporal Action Proposal Combining Watershed and Regression Networks[J]. Chinese Journal of Lasers, 2019, 46(11): 1109001
    Download Citation