Show HN: We collected detailed annotations for text-to-image generation