Special Topics in Multi-Source Feature Fusion
|
App. Res. of Comp. Printed Article
Vol. 43, 2026 No. 3
689-695

Enhanced CLIP semantic features for image captioning

Gu Jinjing1a
Qin Tianbao1a
Pu Yuanyuan1b,2
Zhai Yiqiao1a
Zhao Zhengpeng1a,3
1. a. School of Information Science & Engineering, b. School of Engineering, Yunnan University, Kunming 650500, China
2. The Universities Key Laboratory of Internet of Things Technology & Application, Kunming 650500, China
3. Yunnan Key Laboratory of Low-Light Night Vision Technology and Intelligent Visual Navigation, Kunming 650500, China
DOI: 10.19734/j.issn.1001-3695.2025.08.0270

Abstract

Retrieval-based image captioning leverages rich external textual knowledge. However, existing approaches typically treat the retrieved texts merely as prompts, while relying solely on CLIP visual embeddings for visual information. This app-roach limits performance because the textual prompt information doesn't fully represent the visual embedding space. To address this, this paper proposed the ECSFCap model, an image captioning framework that enhanced CLIP semantic features during encoding and guides the language model to generate high-quality captions during decoding. In the proposed approach, it firstly mapped both textual and visual information into the CLIP space and then transformed them into a multivariate Gaussian distribution. It applied a reparameterization sampling technique to sample from these distributions. Then it processed the sampled results, retrieved texts, and visual embeddings by a cross-encoder to compute the output. This approach effectively combined semantic advantages from both modalities, resulting in enhanced CLIP semantic embeddings. Extensive experiments demonstrate that the semantic enhancement module significantly boosts model performance. Within just 2.1 hours of training, the proposed model achieves a CIDEr score of 122.9% and a SPICE score of 22.5%, making it the fastest model in terms of training time at equivalent evaluation metric levels. These results highlight the model's efficiency and low training cost.

Foundation Support

国家自然科学基金资助项目(52102382)
云南省基础研究计划面上项目(202501AT070234,202401CF070164)
云南省教育厅科学研究基金资助项目(2025Y0008)
云南大学研究生科学基金资助项目(KC-24249132)

Publish Information

DOI: 10.19734/j.issn.1001-3695.2025.08.0270
Publish at: Application Research of Computers Printed Article, Vol. 43, 2026 No. 3
Section: Special Topics in Multi-Source Feature Fusion
Pages: 689-695
Serial Number: 1001-3695(2026)03-006-0689-07

Publish History

[2025-11-18] Accepted Paper
[2026-03-05] Printed Article

Cite This Article

谷金晶, 覃天宝, 普园媛, 等. 基于CLIP语义特征增强的图像描述 [J]. 计算机应用研究, 2026, 43 (3): 689-695. (Gu Jinjing, Qin Tianbao, Pu Yuanyuan, et al. Enhanced CLIP semantic features for image captioning [J]. Application Research of Computers, 2026, 43 (3): 689-695. )

About the Journal

  • Application Research of Computers Monthly Journal
  • Journal ID ISSN 1001-3695
    CN  51-1196/TP

Application Research of Computers, founded in 1984, is an academic journal of computing technology sponsored by Sichuan Institute of Computer Sciences under the Science and Technology Department of Sichuan Province.

Aiming at the urgently needed cutting-edge technology in this discipline, Application Research of Computers reflects the mainstream technology, hot technology and the latest development trend of computer application research at home and abroad in a timely manner. The main contents of the journal include high-level academic papers in this discipline, the latest scientific research results and major application results. The contents of the columns involve new theories of computer discipline, basic computer theory, algorithm theory research, algorithm design and analysis, blockchain technology, system software and software engineering technology, pattern recognition and artificial intelligence, architecture, advanced computing, parallel processing, database technology, computer network and communication technology, information security technology, computer image graphics and its latest hot application technology.

Application Research of Computers has many high-level readers and authors, and its readers are mainly senior and middle-level researchers and engineers engaged in the field of computer science, as well as teachers and students majoring in computer science and related majors in colleges and universities. Over the years, the total citation frequency and Web download rate of Application Research of Computers have been ranked among the top of similar academic journals in this discipline, and the academic papers published are highly popular among the readers for their novelty, academics, foresight, orientation and practicality.

Indexed & Evaluation

  • The Second National Periodical Award 100 Key Journals
  • Double Effect Journal of China Journal Formation
  • the Core Journal of China (Peking University 2023 Edition)
  • the Core Journal for Science
  • Chinese Science Citation Database (CSCD) Source Journals
  • RCCSE Chinese Core Academic Journals
  • Journal of China Computer Federation
  • 2020-2022 The World Journal Clout Index (WJCI) Report of Scientific and Technological Periodicals
  • Full-text Source Journal of China Science and Technology Periodicals Database
  • Source Journal of China Academic Journals Comprehensive Evaluation Database
  • Source Journals of China Academic Journals (CD-ROM Version), China Journal Network
  • 2017-2019 China Outstanding Academic Journals with International Influence (Natural Science and Engineering Technology)
  • Source Journal of Top Academic Papers (F5000) Program of China's Excellent Science and Technology Journals
  • Source Journal of China Engineering Technology Electronic Information Network and Electronic Technology Literature Database
  • Source Journal of British Science Digest (INSPEC)
  • Japan Science and Technology Agency (JST) Source Journal
  • Russian Journal of Abstracts (AJ, VINITI) Source Journals
  • Full-text Journal of EBSCO, USA
  • Cambridge Scientific Abstracts (Natural Sciences) (CSA(NS)) core journals
  • Poland Copernicus Index (IC)
  • Ulrichsweb (USA)