In the fashion sector in particular, it is crucial to manage image data efficiently and consistently. From the correct categorisation and processing of images to the automatic generation of SEO-relevant descriptions – this can now be implemented fully automatically through the use of various AI technologies and intelligent processes – sensibly ‘connected in series’. What this means in concrete terms:
Automatic image role categorisation
To achieve the desired appearance in the web shop, it is essential that each image has the correct image role, is named and tagged accordingly and is stored correctly in the system.
At the heart of the fully automated process for this is an AI Vision component specially trained for fashion images, which analyses images and recognises which of the defined image roles is appropriate. The images are collected fully automatically in advance and status lists are maintained. After the AI Vision analysis, the metadata of the images is updated, image names are modified accordingly and the images are correctly categorised and fed back into the system. In addition, a log is generated to ensure that all processes ran smoothly and all images were processed correctly.
Automatic image processing
Images for the web shop must correspond to the company’s visual language and guidelines. This requires images from different sources to be visually harmonised. Key aspects here are image formats and resolutions, image cropping with defined positioning and white borders, etc. Alignment of articles etc.
To ensure that our customers receive perfect image data for the web shop while minimising the effort required, we have developed specific processes that combine sophisticated scripts with AI services. This means that all of the above can be reliably automated.
Automatic generation of image descriptions, meta tags and SEO texts
In addition to the visual image content, textual elements are essential for online performance. In particular, image descriptions, meta tags and SEO texts are essential. In the past, the creation and management of this textual information was quite time-consuming. In order to be able to realise this highly efficiently for the highest image volumes, we have trained LLMs (Large Language Models) for this purpose and linked them together in a process. This means that all the relevant textual information is now generated on the fly and maintained in the right places. The AI processes also ensure that the texts are directly available in the required and correct HTML formatting.
Interested in innovation too?
Book an appointment with our specialist Christian Weber here.
