Prof. Itti first graduated from the Ecole Nationale Superieure des Telecommunications in Paris before receiving his Ph.D. in computation and neural systems from Caltech. For over a decade, he has been teaching artificial intelligence, vision, and robotics at the University of Southern California (USC, Los Angeles) while publishing abundantly on these topics. But from his experience as a lecturer, he realized that his theoretical courses often lacked an easy to use machine vision component for practical hands-on experiments.
Hence he started JeVois (French for “I see”) as an educational project and the Kickstarter campaign is now turning it into tangible hardware, packing a 1.3Mp video sensor, a quad-core ARM Cortex A7 CPU, a USB video port and a serial port all in a 39x31x23mm self-contained unit including the case and a fan.
According to Itti, it only takes users to load a microSD card with the open-source machine vision algorithms he provides and connect the module to their laptop or Arduino board to give their projects the sense of sight. Video is captured from the camera sensor, processed on the fly through some machine vision algorithm directly on the camera’s own processor, and the results are streamed over USB to a host computer and/or over serial to a micro-controller.
The configurable machine vision engine delivers both visual outputs of how it is analyzing what it sees(so one can understand the algorithms behind vision) and text outputs over a serial link that describe what it has found (useful to send to a micro-controller that can control a robot). Users or machines can also interact with the JeVois smart camera, change its settings, or listen for text-based vision outputs over serial link (both hardware serial and serial-over-USB are supported).
The kit comes with three major modes including a demo/development mode, a text-only mode, and a pre-processing mode which outputs video intended for machine consumption (edge maps or image crops around key features). This pre-processed video can then be further processed by the host computer, for example, using a massive deep neural network running on a cluster of high-power GPUs to recognize the three most interesting objects that the smart camera has detected. Text outputs can be used in this mode too.
While the host computer only has to run a standard video camera software, the camera does all the work, including image capture, vision processing and display results. In fact, the JeVois smart camera can work as a standalone computer, with no USB video streaming. Users could simply stream commands to an Arduino board over the serial port and power the camera via its mini-USB connector.
Get started in machine vision at jevois.org