Spotting all the text in the image with line-level, and output in JSON format. For each text, also detect if it is bold, and estimate font size (small, normal, large).