Artificial Intelligence for the Bin?

AI-supported generation of milieu images for sales and marketing channels – the efficient and effective use of product data is essential for optimally showcasing products. Particularly in digital trade, high-quality product images play a crucial role. That is why Naber, in collaboration with hmmh, has launched a project to generate photorealistic milieu images using artificial intelligence (AI). The aim was to create flexibly deployable image worlds on the basis of a few cut-outs, therefore replacing or supplementing classic photo shoots.

From B2B partner to B2C brand leader: With AI and  flexibility to success

With the acquisition of the trademark and intellectual property rights of Wesco in 2024, Naber was faced with the challenge of preparing a large amount of product data and image material for a variety of sales channels. While Naber had previously operated exclusively in the B2B sector, targeting kitchen studios, furniture stores, and the furniture and appliance industry, the acquisition of Wesco brought a significant change: Wesco's successful B2C shop and strong social media presence made the end consumer the primary target group. This shift required a new approach to product presentation and communication. In particular, the waste bins from Wesco, which play a central role in the product range, needed to be attractively showcased to optimally reach both existing B2B customers and the newly acquired B2C target group.

Since social media and e-commerce platforms continuously require new content, a key requirement was to be able to generate milieu images as flexibly as possible. These needed not only present the products realistically but also depict different application scenarios – be it in the kitchen, bathroom, children's room or other environments. To this end, an innovative solution was developed that uses AI to generate images and can be integrated into the existing PIM system.

From concept to implementation: With the hmmh model to the ideal imagery for B2C

After the successful integration of the data and the
expansion to the B2C target group, it was crucial to optimize the communication and product presentation specifically for this new target group. The visual presentation of the products played a central role in this to reach both the end consumers and the B2B customers in an appealing way. What this looks like in practice and what specific steps were taken for optimization is shown by hmmh's three-stage process model:

1. Proof of Concept:
At the beginning of the project, a real kitchen environment was used and an isolated waste collector was integrated into the image using hmmh's tools. Considering the low level of effort, the results in terms of lighting conditions, reflections and shadows were quite good. However, the size ratio of the waste bin in relation to the height of the worktop was problematic. To further optimize the results, additional media was used and a LoRA (Low-Rank Adaptation) was trained.

2. Training the AI with LoRA models:
The next important step was to train the AI specifically for Wesco products. Although the generative model already had general knowledge about waste bins, it lacked the ability to specifically recognize Wesco products. Training with LoRA enabled the AI to learn to correctly interpret the brand-specific features of the Wesco waste bins and to integrate them realistically into images. Today, Naber achieves significantly better results through the use of AI, especially with regard to shadows, reflections, etc. However, texts and logos still pose a challenge for the AI. But here, too, it would be possible to train a LoRA to show the model what the Wesco logo on the waste bin looks like.

3. Use of 3D models for precision:
In order to further improve the results, the aim is to work with 3D models. Today, 3D models of kitchens are created. Based on these models, new settings can be generated. And with the knowledge of the AI as to the height of a worktop, for example, the waste bin can be integrated into the kitchen model realistically and to scale. The enrichment of further information, such as color codes available in the PIM, leads to ever better images.

Advantages of AI-supported image generation

Thanks to this innovative solution, Naber can now:

• Generate high-quality, photorealistic product images efficiently and cost-effectively
• Flexibly depict different settings and applications
• Ensure the consistency and recognizability of the Wesco brand
• Integrate the content creation process directly into the PIM system
• Reduce dependence on expensive and time-consuming photo shoots

Another advantage: In contrast to conventional generative AI tools, which deliver different results for identical inputs, hmmh's system has been optimized to deliver reliable and reproducible results.

Conclusion

This project impressively demonstrates how the targeted use of artificial intelligence can not only revolutionize product presentation, but also transform a company's entire communication and brand positioning. By integrating Wesco and developing innovative image solutions, Naber has not only optimized internal processes but also established itself as a future-oriented market leader in the B2C sector. With flexible AI-supported image generation, the company has gained a clear competitive advantage and created the basis for sustainable success and brand recognizability – whether in the B2B or B2C sector.