The screenshot from the page itself (the split original/translated) does not translate the image labels. Is that on purpose? It only translates text and does not do OCR?