Description
The Smartpixels project aims to deliver on-device data reduction using neural networks for fine granularity pixel sensors used in high-precision tracking detectors. This has resulted in two major implementations: a filter network and a regression network. Both of these networks deliver novel capabilities for pixels sensors, including on-sensor background rejection and single-sensor reconstruction of the charged particle incident angle. These capabilities are sensitive to the pitch and depletion thickness of the silicon device, as both of these parameters affect the amount of information presented to the neural network. In this contribution we will discuss the performance of the smartpixels regression and filtering neural networks as a function of the device parameters, demonstrating a codesign strategy that covers sensor, readout-asic, and machine learning.
| Focus areas | HEP |
|---|
