RoboScan automatically digitizes analog films with a DSLR camera. A Raspberry Pi is used for this, which uses machine learning and a stepper motor to move the film in a Lego frame. RoboScan is an open source project by Maker Benjamin Bezine that was created in lockdown and is active on GitHub as bezineb5.

During operation, the film is illuminated from behind at the point to be scanned. This classic trick can be used to photograph the film using a camera with a macro lens. A 3D-printed adapter is required so that the stepper motor can move the analog film in the Lego structure. Versions 2, 3 and 4 are suitable as RaspberryPi. The Pi 4 has the advantage here that the camera can be connected via USB 3.0. Bezineb5 published the project and the material list on GitHub,

Software control RoboScan consists of a JavaScript web front end and a Python backend that takes over the control. The scanning process can be started via the web interface and the images saved on the computer – so there is no need to fiddle with the SD card. The software uses the library libgphoto2 to operate the camera.

A Google Coral Edge TPU USB Accelerator can optionally be used to optimize the machine learning component. How the machine learning algorithm recognizes the individual image segments is clearly explained in the following YouTube video. The installation runs via a supplied Docker-Compose-File, which starts all components and connects them with each other.

And this is how the RoboScan looks during operation.

(stri)