Home Artificial Intelligence Exploring the chances of Generative AI for Real Estate: use cases at Casavo.

Exploring the chances of Generative AI for Real Estate: use cases at Casavo.

Exploring the chances of Generative AI for Real Estate: use cases at Casavo.

Variations of a Bedroom generated using Stable Diffusion — Credit

Casavo is on a journey to alter the best way people sell, live, and buy homes in Europe. Our modern approach to real estate is disrupting the normal market by leveraging advanced technology to streamline the method and make it more transparent and efficient for everybody involved.

To attain this goal, we harness the ability of Machine and Deep Learning techniques to tackle quite a few challenges in the true estate industry. With AI, we are able to predict the worth of a property and analyse home images to realize insights into their contents, anonymise them, and detect NSFW material to share images safely with potential buyers. Nonetheless, with the recent advent of generative AI, we cannot help but consider the potential applications of this technology in our company. That’s why we conducted a study to explore how diffusion models could speed up a few of our internal processes and enhance our services even further.

Images Generated by Stable Diffusion — Credit

Generative AI is a fast-developing field that utilises deep learning models to create recent data, including images and text. Over the past two years, we’ve witnessed a surge in using these models and their increasing presence in our lives. Notably, the emergence of ChatGPT, which took just two months to succeed in 100 million users, has demonstrated the potential of AI-generated text. Similarly, advancements in image-generation models like Midjourney, DALL-E, and Stable Diffusion have shown that these models are capable of making images with a high degree of fidelity and creativity.

Amongst the various generative AI models available, Stable Diffusion has emerged as a promising model that would potentially help Casavo take its services to the subsequent level. This model, developed by the startup Stability AI, relies on the principles of diffusion processes in physics and arithmetic. It operates by iteratively refining an initial noise signal to generate images with a high degree of realism and creativity. What’s great about this model is that the denoising process, which results in the generation of the image itself, could be guided using a textual prompt. These features make the model versatile and suitable for various tasks, including text-based image generation, image inpainting, image-to-image variation generation, and super resolution.

Image Inpainting using Stable Diffusion — Credit | Credit

For example, these images were created using Stable Diffusion and text-guided inpainting, starting with a dog on a bench and the famous Windows XP wallpaper, respectively. In each images, we masked a portion that was then translated into noise. We then let Stable Diffusion remove the noise while guiding the method with textual prompts, similar to “some sheeps on the grass”

In case you are interested by the inner workings of Stable Diffusion, this great article titled “The Illustrated Stable Diffusion” by Jay Alammar is a very good place to start out.

By incorporating generative AI into its operations, Casavo can improve its internal processes and supply clients with a more immersive and interactive experience. The chances of using generative AI in our business are quite a few, and we’ll explore among the most promising ones here, including property renovation, organic traffic generation, and property image enhancements. Let’s take a better have a look at each of those use cases and supply some examples.

One among the important thing facets of our iBuyer business model is acquiring properties, renovating them, after which selling them in the marketplace. Nonetheless, the technique of renovating an apartment and furnishing it to appeal potential buyers is a labor-intensive and time-consuming task. By utilizing generative AI, we are able to potentially speed up this process and boost the creativity of our team.

With the assistance of Stable Diffusion, we are able to select an image of an apartment and generate a renovated version of the property with a very different style and furniture. By adjusting the quantity of noise added to the unique image, we are able to control the extent of creativity within the output and the preservation of the unique key features. As an instance this, we now have generated some variations of a property by adding 25% noise to one in every of its images and used different prompts to guide Stable Diffusion in imagining the identical property renovated in multiple styles. Listed here are the outcomes.

An image of a bedroom in an apartment — Credit
Image and Noise combined are fed to Stable Diffusion, with guidance text prompts
Stable Diffusion de-noises the input image following the textual guidance

As demonstrated, generating multiple variations of a picture with different and attractive furnishing styles is comparatively straightforward. Probably the most significant effort is required within the text prompt engineering, which might produce vastly different results. We needed to rigorously adjust each the quantity of noise added to the input image and the text-guidance weight while generating the outcomes to acquire relevant outcomes. It is important to notice that, despite its potential advantages, Stable Diffusion technology is just not yet advanced enough to completely replace the manual technique of renovating a house, including choosing and arranging furniture that was not present during training. Nevertheless, this technology can undoubtedly aid our team in generating creative ideas to renovate apartments.

One other use case pertains to how this technology could potentially enhance the organic traffic for buyers on our listing platform. For example, buyers could use this technology to visualise how an empty apartment on the market would seem like with furniture just by invoking Stable Diffusion through an internet interface. This feature might bridge the gap between viewing an inventory and visiting the property, because it allows users to assume how an empty space could seem like with furniture. This could be possible by the Stable Diffusion inpainting functionality, which translates to adding 100% noise to a portion of the image (a mask) and have the model reconstruct it from scratch using a text prompt as guidance.

While experimenting, we pretended to have this tool already deployed and tested it on a property on the market on our platform. We masked portions of the space and used text-prompts so as to add furniture pieces at each iteration. The outcomes are impressive, and we would love to share a few of them with you.

An empty listing on the market on our platform

We chosen one in every of the images of this lisiting and iteratively masked a portion of it, added a prompt, and asked the model to inpaint with recent furniture. We briefly highlight the method for a single prompt, after which showcase a collage of images with a summary of every step.

Means of Inpainting using Stable Diffusion with prompt “An empty kitchen”Credit

As you’ll be able to see, the model flawlessly performs inpainting on the masked area of the image, reconstructing the background scenery while retaining the unique blur effect and erasing the person. Undoubtedly, this method can prove useful in various other contexts, which we will explore later. Nevertheless, for this particular use case, we are going to utilise inpainting solely to enhance the looks of a property by adding furniture, as demonstrated in the following pictures.

Iterative Image Inpainting to furnish an empty room

And identical to that, we generated two easy concepts for furniture in our apartment (please don’t judge my design skills 😅). A possible buyer could rejoice using this tool on our platform if it’s served in an intuitive and user-friendly manner. In the long run, we could even mix the previous ideas to generate variations of the apartment furnished by the customer, after they’ve selected the important thing facets they’d wish to see within the room. Sticking with the Scandinavian and Industrial styles, these images depict possible variations which preserve the concepts of a room furnished with a bed, a carpet and a painting.

Scandinavian and Industrial variations of a furnished room

Lastly, this technology could be useful for robotically anonymising pictures or documents that we collect from our clients for downstream tasks, similar to showing apartment previews to potential buyers. On this scenario, we are able to use a straightforward pipeline with object or text detectors to robotically mask “sensitive” portions of the image. With the assistance of Stable Diffusion, we are able to then ask the model to reconstruct the image without the sensitive content. Nonetheless, this approach has the disadvantage of doubtless generating artefacts in the pictures, so it is just not yet reliable enough to be implemented in production. Future versions of such generative models may allow for this task to be carried out autonomously. Currently, we don’t integrate Stable Diffusion into our anonymisation pipelines and like to blur out content flagged as sensitive or NSFW. No matter this, listed here are a couple of images that depict what we could possibly construct in the long run.

Possible Stable Diffusion Anonymisation Pipeline — Credit

Impressive, isn’t it? It’s value noting that every one of those images were processed using Stable Diffusion, somewhat than being manually edited with tools like Adobe Photoshop. This technology could prove to be extremely useful not just for actual images, but in addition for documents similar to floor plans. For example, the next image demonstrates how our anonymisation pipeline detects and removes text. Nonetheless, as a consequence of the technique we use (blurring and color alternative), the outputs could be quite noisy. By implementing Stable Diffusion, we could potentially improve this anonymisation process even further.

Our current document anonymisation pipeline, without SD. As you’ll be able to see from the rightmost image, results are quite noisy. SD could improve the outputs. (Sensitive information have been manually brushed away with an orange brush within the leftmost image for obvious reasons)

In conclusion, Casavo’s commitment to revolutionising the true estate industry using advanced technology is clear in our use of Machine and Deep Learning to simplify the technique of buying and selling a house. Nonetheless, we’re not content with stopping there. Generative AI is an exciting and promising field that would enhance Casavo’s services even further. Stable Diffusion, specifically, has shown remarkable ends in creating images which can be each realistic and inventive. Incorporating generative AI into our operations could improve our internal processes, provide clients with a more interactive experience, and assist our internal employees. While generative AI may not replace manual work entirely, it has the potential to generate creative ideas and expedite processes.

Casavo is all the time on the lookout for modern ways to enhance our services and supply our clients with one of the best possible experience. In case you share our vision for transforming the best way people sell, live, and buy homes, we invite you to try our open positions and join us on this exciting journey. Together, we are able to create a more transparent and efficient real estate marketplace using cutting-edge technology 🏠 🚀



Please enter your comment!
Please enter your name here