Скачать или смотреть Keynote On-Sensor Computer Vision With Pixel Processor Arrays Workshop on Neuromorphic Vision ICCV25

Keynote: "On-Sensor Computer Vision With Pixel Processor Arrays, Current and Future Opportunities"

2nd Workshop on Neuromorphic Vision NeVi ICCV 2025, Honolulu, Hawaii, United States, 20th Oct 2025

Prof Walterio Mayol-Cuevas

Work with Prof Piotr Dudek, Dr Laurie Bose, Hongyi Zhang, and colleagues from Bristol and Manchester Universities.

Abstract: On-sensor Computer Vision offers opportunities that have seldom been available to artificial systems before. Low-latency visual computation, reduced power and space budget and low bandwidth requirements are at the centre of the challenges preventing efficient visual systems to scale and be deployed in the wild. If images do not need to be ferried around or reconstructed before understanding, multiple applications from robotics to IoT to all-day wearables will be unlocked. Parallel processor arrays (PPAs) are a new class of vision sensor devices that exploit advances in semiconductor technology, embedding a processor within each pixel of the image sensor array. Sensed pixel data are processed on the focal plane, and only a small amount of relevant and already processed information is transmitted out of the vision sensor. This tight integration of sensing, processing, and memory within a massively parallel computing architecture leads to an interesting trade-off between high performance, low latency, low power, low cost, and versatility in a machine vision system. In this talk, we will cover recent research that showcases a range of visual competences and applications achievable with on-sensor computation as well as introduce the challenges that a new research project between the Universities of Manchester, Bristol and Imperial College is aiming to tackle.