The rapid development of AI has brought about continuous breakthroughs in the multi-modal of large models. Among them, the multi-modal technology of generative tasks represented by AI image generation has gradually matured, promoting the development of the geographic spatial planning and design industry. It also indicates the new stage of intelligent design.
Geospatial design adopts advanced concepts of spatial design and combines geographic spatial information to complete the design of two- and three-dimensional urban space, urban environment, building layout and buildings image. It provides creative and desirable spatial planning.
In the traditional geospatial design process, there are many time-consuming manual tasks such as the generation of initial concept maps, the implementation of preliminary plans, and the modification of following designs. Using traditional design software to carry out geospatial design means incurring huge manpower, time and budget costs.
Based on large models and multimodal technologies, SuperMap has developed an industry application of AI image generation, which can generate high-quality renderings in batches within seconds, helping designers shorten the design process and solve problems such as inefficiency and time consumption. It can effectively form new quality productivity and empower geographic spatial design.
01 Typical functions of AI image generation application
The AI image generation application is the application of multimodal technology in vertical industries. Its main functions cover three major directions: AI image generation, secondary editing, and model management, which can help improve quality and efficiency. The following will introduce several typical functions in detail, including image generation by text, image generation by image, partial redrawing, super-resolution reconstruction, image generation by vector extraction, and generated image style fine-tuning.
Image generation by texts
As the most classic function of AI image generation, image generation by text can help designers quickly implement their design ideas. Based on this function, designers only need to enter the text descr1ption of the desired image and wait for a few seconds to get multiple high-quality images with diverse design styles, which will greatly inspire designing. SuperMap AI image generation application provides a massive prompt word library for planning industry, and you can generate high-quality concept images with a simple click.
As shown in the following examples, through the simple descr1ption like "commercial building, evening, glass curtain wall" combined with the "commercial building" style, conceptual maps of commercial blocks in the evening can be quickly generated in batches.
Image generation by images
The controllability of image generation is a key requirement for industry applications. In scenarios where accurate image generation is required, the reference image constraints can meet the needs of both creativity and accuracy. Controlled image generation can simultaneously support single or multiple constraints based on the line draft, outline, depth, line segment, semantics, graffiti, color, and light and shadow of the reference image to achieve controllable output effects.
In the picture below, the overall structure of the building is constrained by line drawings, and the ambient light and shadow are constrained by glass curtain wall buildings. The two are combined to generate an architectural design drawing that takes into account both structure and light and shadow.
Single constraint control image generation
Multi-constraint control image generation
Partial redrawing
Partial redrawing is used for automatic image repair and intelligent image modification. The redrawing area can be customized. The AI algorithm automatically redraws according to the picture descr1ption, accurately repairs and optimizes the partial area in the picture, and provides designers with a variety of repair results for reference, making creation easier.
In the picture below, the high-rise building that needs to be modified is partially painted and redrawn to generate a new high-rise building that conforms to the picture structure of the original image.
Partial redrawing
Super-resolution reconstruction
Super-resolution reconstruction technology converts low-resolution images into high-resolution versions through generative AI amplification algorithms and detail enhancement, significantly improving the clarity and detail of the image. It is an important post-design process.
In the figure below, the low-resolution, low-quality blurred image generated in the previous stage is processed using super-resolution reconstruction technology to generate a high-quality image after high-definition enlargement.
Examples of super-resolution images
Vector extracted image generation
After the introduction of AI mapping technology, generating urban planning concept maps can be completed in only three steps. First, use the AI model to automatically extract the content that needs to be generated (such as buildings, etc.) from the floor plan, and then intelligently stretch the height of the building to obtain a three-dimensional image of white model at a certain perspective, and finally generate a three-dimensional scene image of a certain perspective and style through constrained rendering. The entire process does not require modeling expertise, and the design threshold is lowered.
As shown in the following figure, the building roof vectors are extracted from the master plan, and then the 3D white model of the building is stretched. After the 3D white model image is intercepted, the 3D scene concept map of the area is generated through constrained image generation.
Example of image generation by vector extraction
Image style fine-tuning
How to generate pictures with exclusive styles for specific scenarios? For example, in the scenario of rural renovation design, more rural factors are needed; when planning the master plan design, a standard color scheme is needed, etc. AI image generation provides exclusive style fine-tuning capabilities, and only a small amount of picture materials can be used to create a fine-tuning model with user-customized styles.
In the example below, a small number of commercial building images can be used to generate a unique commercial building style fine-tuning model to use in generating images after a short period of training with low threshold. At the same time, the product has built-in 8 commonly used industry style fine-tuning models, and supports users to upload local style fine-tuning models for use.
Customize your own image style
Examples of some built-in exclusive image style models
02 Application scenario display
The spatial planning AI image generation application help accelerate the spatial planning and design process. From concept generation to detail optimization, it provides a variety of image generation modes, covering most scenarios in the field of spatial planning and design, providing more efficient solutions for planning, urban design, spatial transformation and other scenarios. Its main application scenarios include:
Quick implementation of concept design
The application can automatically generate planning and design concept drawings. It provides a variety of styles and options to choose, helping designers quickly determine the design direction and style in the early stages. It helps improve the efficiency at the early design stage, and accelerates the process of concept verification.
Precise design and intelligent planning
AI image generation constraints generation results. Sketches can be converted to finished renderings with one click, and rendered in proportion. Architectural and landscape planning and design plans of different styles and functions can be intelligently generated to simulate building layouts, and optimize greening configurations.
Generation of high-quality concept maps for fixed-view 3D scenes
AI generation can be used to generate high-precision and efficient 3D scene images. At the same time, spatial layouts under different environments and lighting conditions can be quickly simulated to display the overall design details and realistic effects of buildings and landscapes.
Spatial Planning Style Migration
Customize your own style through the application. You can migrate the existing space planning and design style to the new plan to quickly generate design renderings with a unified style.
Optimization of details
The application can improve the functionality and aesthetics of building images without changing the overall design. It automatically handles tedious and repetitive design tasks through image redrawing, repair and fine-tuning.
The application of AI image generation for spatial planning can not only improve the design efficiency of geospatial designers, but also enhance the creativity and flexibility of design works. Designers can easily generate and optimize design plans, quickly respond to customer needs, give spatial planning and design greater possibilities and innovation space, and bring a new perspective and direction to the planning and design industry. In the future, with the continuous advancement of AI technology and cross-industry integration, AI image generation is expected to achieve creation with higher precision, larger scale, and more professionality.
