[Siva Shanmugam Manickam]

I am currently developing a platform that provides home decor design suggestions. The objective is to create personalized home decor designs by incorporating images of my own furniture into users’ style preferences, like ‘Modern Minimalist’. For instance, if the user selects a modern minimalist style with a gray sofa and a glass coffee table, the platform should generate a design using images of these specific furniture pieces.

I understand that Large Language Models (LLMs) like Stable Diffusion are trained to generate such home decor designs. However, I am seeking clarification on the most effective strategy from both cost and performance to realize this vision. I’m considering whether to fine-tune the Stable Diffusion model with images of my furniture or use techniques such as inpainting (though I’m uncertain if this is the correct approach), or perhaps another method entirely.

I am looking forward to any suggestions and recommendations from the members of this forum to help guide me in the right direction

Thanks,

Siva Manickam

• This topic was modified 11 months ago by Siva Shanmugam Manickam.

[Andrew]

You can first start with prompting to see if you are about to get you want. Sdxl model is a good starting point.

you can train a Lora if you want to incorporate your own furniture.

[Siva Shanmugam Manickam]

Thank you, Andrew, for your quick reply. I understand that while Prompt can create designs, it might not be able to generate them using my images. Is this correct, or is there a method within Prompt to create home decor designs incorporating my images? Regarding the training process, you’ve suggested using the SDXL model with LORA. I will begin my research on this and reach out if I have any questions

Thanks,

Siva Manickam

 

[Andrew]

That’s correct.

If you want to generate furniture with a similar style, you can use an IP-adapter controlnet and supply your images.

https://stable-diffusion-art.com/image-prompt/

[Author]

Posts