A new paper is out: All-Analog Silicon Integration of Image Sensor and Neural Computing Engine for Image Classification, a collaboration among Università della Calabria, Quantavis and Università di Pisa.
A single silicon chip demonstrator including image sensors and a low-resolution image classification neural network, all based on analog computing and signal processing on standard CMOS technology.
The rationale is that if we can suppress energy- and time-inefficient operations, such as analog-to-digital conversion and data transmission, we can build a cognitive image sensor, with embedded classification capabilities for applications in which low power and small volume are critical, such as wearable and mobile healthcare electronics or autonomous microrobots and drones.
In parallel, we investigate this concept for a technology based on 2D Materials in the H2020 QUEFORMAL project (Quantum Engineering for Machine Learning https://www.queformal.eu) , and for sensors in harsh environments using silicon at very high temperatures in the Charm ECSEL project https://charm-ecsel.eu/project/.
Kudos to Benjamin Zambrano, Sebastiano Strangio, Marco Lanuzza, Tommaso Rizzo, Esteban Garzón.
By the way, next week we start teaching a full new course on neuromorphic circuits and devices
full paper here: https://ieeexplore.ieee.org/document/9874841