Hi, I wrote a blog post about the detection of digits on a display. It's written in french (sorry!) there https://blog.sciunto.org/posts/detection_nombres/ but the snippets talk by themself. My point is that it could be a pain to collect data from electronics (no serial port...) and displays are everywhere, camera largely available in labs. Basically, the idea is to use scikit-image and then scikit-learn. Scikit-image to isolate digits and make sub-images with each of them. Scikit-learn to recognize the digits. A training of 3 example per digits gave me a 100% success over 60 images. At the bottom, I show that it is also robust to the superposition of two digits (because of a fluctuation during the snapshot). Best, -- François Boulogne. http://www.sciunto.org GPG fingerprint: 25F6 C971 4875 A6C1 EDD1 75C8 1AA7 216E 32D5 F22F