Work

Contrastive language-image pre-training (CLIP) in e-commerce: applications, methodologies, and performance

Year: 2025

Type: article

Abstract: This article thoroughly examines the architecture and applications of the Contrastive Language-Image Pre-training (CLIP) model within the e-commerce domain, focusing on key tasks such as visual search, product recommendation, and attribute extraction. The article also provides an in-depth analysis of the methodologies used for CLIP’s adaptation to e-commerce tasks and the relevant datasets employed. By highlighting the unique capabilities of the CLIP model, such as its ability to perform zero-shot learning and contrastive pre-training, this article underscores its potential impact on the industry while also acknowledging its limitations, including the ‘domain gap’ and the need for adaptation strategies. Furthermore, the article explores the future research directions for enhancing CLIP’s performance in specialized e-commerce contexts and compares it with other traditional and multimodal AI techniques.

Source: Advances in Cyber-Physical Systems

DOI: 10.23939/acps2025.01.100

Authors: Oleksandr Khainas, Nataliia Melnykova, Solomiia Fedushko

Institutions: Lviv Polytechnic National University, University of Agriculture in Krakow, Comenius University Bratislava

Language: English

Topic: Multimodal Machine Learning Applications

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

Open Access status: bronze

OpenAlex ID: W4411230698

FWCI: 0.374

Cites: 12

Cited by: 1

Related to: 10