In-sensor computing allows image sensors with internal computing capabilities to reduce communication latency and power consumption for machine vision in distributed systems and robotics. Because of its tunable electrical and optical properties and amenability for heterogeneous integration, two-dimensional semiconductors have several advantages in creating intelligent vision sensors. Researchers have developed a multipurpose infrared image sensor based on an array of black phosphorus programmable phototransistors (bP-PPT).
It is possible to remotely program the electrical conductance and photoresponsivity of the bP-PPT locally or remotely with 5-bit precision. The execution includes changing the stored charges in the gate dielectric layers electrically and optically to create an in-sensor convolutional neural network (CNN).
The sensor array can receive optical images broadcast in the infrared over a wide spectral range and process and recognize them with 92% accuracy using inference computation. The black phosphorus image sensor array can be scaled up to create a more complicated vision-sensory neural network, which has a lot of potential for distributed and remote multispectral sensing applications.