Imagine uploading an image into the system, which in turn analyses which item is depicted and automatically assigns it the appropriate customs tariff number. Sounds nice, doesn’t it? Automatic image recognition would completely revolutionise the customs sector. But how could this work, and how far are we from implementation?