SRT2ch: Large-Scale Human Image Editing Dataset

This project addresses a critical gap in image editing AI: the lack of datasets reflecting real human editing preferences across different cultures. By collecting and analyzing 10,000+ authentic image edits from Chinese and Japanese social media platforms, we aim to train more culturally-aware diffusion models. The research tackles several challenges: - Privacy-preserving data collection methodologies - Cross-cultural aesthetic preference analysis - Ethical considerations in scraping social media - Training diffusion models on diverse editing styles