The $300 system in the fight against illegal images - BBC News
▻https://www.bbc.com/news/technology-44525358?ocid=socialflow_twitter
Mr Haschek used three Raspberry PIs, powering two Intel Movidius sticks, which can be trained to classify images. He also used an open source algorithm for identifying explicit material called NSFW (Not Safe For Work), available free of charge from Yahoo.
Image copyright Christian Haschek
Image caption Christian Haschek said the system took a couple of hours to assemble.
He set it to find images which the system could say with 30% or more certainty was likely to contain pornography - he said he set the possibility low so as to be sure not to miss anything.
He has since discovered 16 further illegal images featuring children on his platform, all of which he reported to Interpol and deleted.
He then contacted a larger image hosting service, which he declined to name, and found thousands more by running images uploaded to their platform through his system as well.