SSG.com equips 'SseukLens' with multimodal AI... First in domestic e-commerce

Simultaneous Learning of Product Image and Text Information

SSG.com announced on the 26th that it has applied its self-developed multimodal artificial intelligence (AI) model, the first of its kind in the domestic e-commerce industry, to its deep learning-based image product search service, 'SSG Lens.'

SSG Lens is an image search service operated by SSG.com since 2018. When users press the lens button on a product detail page, it recommends products similar to the one currently being viewed. By clicking the lens icon to the right of the search bar and uploading a photo via smartphone, it recommends similar products based on the product’s appearance, attributes, color, and more. Multimodal AI refers to a model trained on both product image information and textual information such as name, brand, and detailed features. With this update, a beta version of multimodal AI has been added to SSG Lens, enabling it to process complex information combining images and text.

Previously, SSG Lens could only search for similar products using image information alone, but with the application of multimodal AI, the 'Add Search Term' feature allows users to input text alongside image search results. This upgrade aims to provide customers with customized search results tailored to their preferences.

For example, after searching with an image of a specific clothing item, users can narrow down their search by entering additional keywords related to product attributes such as color (beige) or design (flower pattern) in text form. SSG.com explained that the AI enhancement has significantly improved image search and classification capabilities compared to before, allowing users to quickly compare similar products through recommendations, thereby reducing shopping time. Additionally, style- and brand-based products related to the image are also recommended.

Jang Wonju, head of data at SSG.com, said, “SSG Lens is the first case of applying the multimodal AI, which we have been developing and using for AI training since early last year, to customer service.” He added, “We plan to expand the use of multimodal AI to areas such as product detail pages and search results for search and recommendation purposes in the future.”

This content was produced with the assistance of AI translation services.