Attention based CNN-LSTM network for video caption

XinBo Ai; Qiang Li

doi:10.1117/12.2652227

10 November 2022 Attention based CNN-LSTM network for video caption

XinBo Ai, Qiang Li

Proceedings Volume 12331, International Conference on Mechanisms and Robotics (ICMAR 2022); 123315O (2022) https://doi.org/10.1117/12.2652227
Event: International Conference on Mechanisms and Robotics (ICMAR 2022), 2022, Zhuhai, China

Abstract

Due to the demand and wide application of video caption in various fields such as video retrieval, content recommendation, risk management, etc., how to extract a comprehensive and highly generalized description of the information has been an active research area for many years. In this paper, we propose a new model that includes the fusion of convolutional neural network and attention mechanism. The special features are extracted from multiple perspectives such as scene, target and behavior in the video, and combined with key frame semantic information to reduce the interference of redundant information and complete the feature representation of the information, and the attention-weighted fusion of the input of the above four features is input to LSTM decoding to finally generate the video content title. A multi-baseline comparison of the two public datasets is performed. Multiple evaluation metrics prove that our model outperforms other models and also show that the information representation of this paper.

Citation Download Citation

XinBo Ai and Qiang Li "Attention based CNN-LSTM network for video caption", Proc. SPIE 12331, International Conference on Mechanisms and Robotics (ICMAR 2022), 123315O (10 November 2022); https://doi.org/10.1117/12.2652227

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available