Technically this wouldn’t be that hard. As mentioned, a python/ruby/whatever script could do it by making use of the “Download unprocessed original” link that is only available to the user who uploaded an image. However the Mapillary terms of use do explicitly prohibit “mass downloads or bulk feeds of any Content, including Metadata”
Of course I have no idea how strictly that clause would be enforced if you were only downloading your own images. It might depend on how many you have. I’m pretty sure trying to download 500k images in one go would get you a pretty quick ban 