Head to the "Launch URL" (https://a-eye-vision.tech/ - a hosted version of https://github.com/A-Eye-Project-for-CSC1028/a-eye), and either use the provided low-poly 3D model, or upload your own via the "Select Model" button. Toggle depth map, and tweak settings as required to get a balanced depth map. Then, position the camera as you like - hit "Render Image" to save out the depth image for use with the generation script. Git clone the following: https://github.com/A-Eye-Project-for-CSC1028/a-eye-generator - ensure you also have ComfyUI installed with the StabilityAI control-lora package (https://huggingface.co/stabilityai/control-lora) and a Stable Diffusion model of your choice. I used SDXL (https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0), and you can do the same. If you use something else, additional configuration may be required. In the root of the newly-cloned repo, change the value of "COMFY_DIRECTORY" in .config to be the root of your ComfyUI installation. If you used a different control-lora or model than me, please open "generate.py" within scripts/image-generator and change the relevant values as commented. Finally, install all relevant packages, write the main.py to your liking (import generate(), GenerationParameters, etc.), and run! This has been an awesome month of building; I didn't quite get as much done as I would have liked due to circumstances outside of my control, but I am really thankful for the opportunity! Maybe I'll join V4 or later and try something else next time...
