Exploring AI Image Generation for Landscape Architecture Designs
| Category | AI image generation Developing AI literacy |
| Instructors | Jennifer Birkeland, Assistant Professor |
| Landscape Architecture | |
| College | College of Agriculture and Life Sciences |
| Course | LARCH 5060: The Future of Image: Tools in Representation |
| Discipline | Landscape Architecture |
| Course-level | Graduate |
| Course size | 20 students |
| Implemented | Spring 2024 |
Learning Outcomes Targeted
AI Literacy and Proficiency
Problem-based Learning
Visual Communication
Refining AI-generated Images
Brief Summary
Students used a variety of prompts with AI image generators to create a large volume of sample images to bring to class. In class discussions, students looked critically at the limitations of AI based on the images created from their prompts. The discussions covered such topics as biases, ownership, and the source libraries for the AI tools. The students then chose one of their prompts and an associated image to refine into a 3D model crafted by the student.
Context
In the field of landscape architecture, designers rely on images and 3D models to communicate with clients. AI image generation technology will almost certainly be part of future work in making images from design ideas. The instructor decided to dive in alongside her students to test how well these tools work. As a class, they went back and forth between using AI and other methods of creating drawings and models, exploring their potential uses and limitations.
Did the prompts match their expectations?
What were the shortcomings?
What biases existed within the AI databases?
Implementation
Using AI image generators such as DALL-E2 and MidJourney, students were tasked with bringing in 80 images of a “weird thing,” focusing on how the manipulation of their prompt could alter the image. This creation was followed by a class discussion: Did the prompts match their expectations? What were the shortcomings? What biases existed within the AI databases?
Following this exercise, students chose one of their prompts and evaluated the image outputs of that prompt based on criteria discussed in class. After a deeper evaluation of the images, they chose one of these creations and turned it into a 3D model to further meet the course’s core learning objectives.
Challenges
The initial challenge that students encountered was a lack of familiarity with the AI tools. Through continual exposure and by Birkeland creating a safe space for students to explore, they were able to overcome this challenge. The larger challenge discovered through the project was the realization of the limits of AI libraries, specifically regarding design and style. The students found that including culturally loaded content into the prompt could confuse the AI resulting in what they referred to as “blobitecture”: completely useless pieces fused together in no recognizable pattern.
“It’s helping my less creative students be more creative or think outside the box a little bit more. So that’s where I definitely appreciate it.”
Jennifer Birkeland
Reflection and Future Directions
At this time, the instructor believes that the incorporation of more AI animation earlier may be beneficial. Building students’ confidence in using the methods over time can help ease them into using these tools
How to Implement This in Your Class
Some key tips to keep in mind when working with AI image generation are:
- Try to understand the models/libraries/sources that AI uses. Recognizing the limitations, biases, and information that the AI uses when creating the images is crucial. Use this as a point of discussion with your students to help them understand AI more and be more critical of its use.
- Start small and build. You do not need to change your entire syllabus to incorporate AI into your class, as this can be overwhelming. Instead, focus on one assignment, activity, or discussion where you can have the students use AI. From that point, be critical of how that instance went and if it’s beneficial to include it more in your course
- Create a safe space for exploration and discussion within the class. Give the students a simple prompt to start with for the generator, and then encourage the students to bring in a multitude of results. Giving students the ability to explore the variances that can exist with a simple prompt and by manipulating it can establish a great foundation for a discussion.
- Make sure to give time and space for students to reflect on their use of AI image generation based on their experiences in class to help them further their understanding.