Optics and Precision Engineering, Volume. 30, Issue 24, 3198(2022)
Chained semantic generation network for video captioning
Figures & Tables(15)
Fig. 2. Diagram of visual feature segmentation and overall network structure
Table 1. Performance comparison of ablation experiments
View tableView in ArticleTable 1. Performance comparison of ablation experiments
MSR-VTT MSVD Mode B-4 C M R B-4 C M R G 45.83 53.16 29.28 63.64 62.35 109.71 39.04 77.04 L 45.21 52.63 29.38 63.59 61.07 113.41 39.71 77.34 G-L 46.70 52.93 29.55 64.03 65.20 119.58 41.51 79.02 交叉 47.44 52.98 29.75 64.11 66.29 119.93 42.25 79.30 非交叉 46.72 52.67 29.60 63.99 65.47 120.09 41.49 79.24 k=1 46.70 52.93 29.55 64.03 65.20 119.58 41.51 79.02 k=2 47.44 52.98 29.75 64.11 66.29 119.93 42.25 79.30 k=3 47.70 53.64 29.78 64.31 65.62 122.44 42.25 79.93 k=4 48.20 53.75 29.98 64.48 66.51 122.26 41.98 80.05 k=5 48.35 54.22 29.90 64.47 64.45 121.09 42.13 79.85
Table 2. Performance comparison on MSR-VTT dataset
View tableView in ArticleTable 2. Performance comparison on MSR-VTT dataset
Method B-4 C M R MTVC ［ 20 ］40.8 47.1 28.8 60.2 CIDEnt-RL ［ 21 ］40.5 51.7 28.4 61.4 SibNet ［ 22 ］40.9 47.5 27.5 60.2 HACA ［ 23 ］43.4 49.7 29.5 61.8 TAMoE ［ 24 ］42.2 48.9 29.4 62.0 POS ［ 25 ］41.3 53.4 28.7 62.1 MARN ［ 10 ］40.4 47.1 28.1 60.7 JSRL-VCT ［ 26 ］42.3 49.1 29.7 62.8 GRU-EVE ［ 27 ］38.3 48.1 28.4 60.7 STG-KD ［ 28 ］40.5 47.1 28.3 60.9 SAAT ［ 29 ］39.9 51.0 27.7 61.2 ORG-TRL ［ 9 ］43.6 50.9 28.8 62.1 SAVC ［ 8 ］45.8 53.2 29.3 63.6 ChainS-Net 48.2 53.8 30.0 64.5
Table 3. Performance comparison on MSVD dataset
View tableView in ArticleTable 3. Performance comparison on MSVD dataset
Method B-4 C M R LSTM-E ［ 30 ］45.3 31.0 h-RNN ［ 31 ］49.9 65.8 32.6 aLSTMs ［ 32 ］50.8 74.8 33.3 SCN ［ 14 ］51.1 77.7 33.5 MTVC ［ 20 ］54.5 92.4 36.0 72.8 ECO ［ 18 ］53.5 85.8 35.0 SibNet ［ 22 ］54.2 88.2 34.8 71.7 POS［ 25］ 53.9 91.0 34.9 72.1 MARN ［ 10 ］48.6 92.2 35.1 71.9 JSRL-VCT ［ 26 ］52.8 87.8 36.1 71.8 GRU-EVE ［ 27 ］47.9 78.1 35.0 71.5 STG-KD ［ 28 ］52.2 93.0 36.9 73.9 SAAT ［ 29 ］46.5 81.0 33.5 69.4 ORG-TRL ［ 9 ］54.3 95.2 36.4 73.9 SAVC ［ 8 ］62.4 109.7 39.0 77.0 ChainS-Net 65.6 122.4 42.3 79.9
Table 4. Runtime comparison
Tools
Get Citation
Copy Citation Text
Lin MAO, Hang GAO, Dawei YANG, Rubo ZHANG. Chained semantic generation network for video captioning[J]. Optics and Precision Engineering, 2022, 30(24): 3198
Paper Information
Category: Information Sciences
Received: May. 16, 2022
Accepted: --
Published Online: Feb. 15, 2023
The Author Email: GAO Hang (gao_hang2021@163.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20