AIGC Scene Design Project—AI Image Generation and Labeling
In the AIGC Scene Design Project, used Stable Diffusion and ControlNet to generate anime-style scenes from line art images. Controlled and structured the output through integrated Depth and Lineart cues to produce consistent, high-quality image data. Developed and iterated workflows in ComfyUI to enhance both efficiency and output quality for generative AI training. • Generated images were labeled and structured for style, composition, and scene attributes. • Performed scene classification and content segmentation to enrich datasets for AIGC. • Focused on sci-fi and anime scene domains to ensure domain specificity of training data. • Integrated workflow improvements designed to support scalable and reliable data annotation.