Search papers, labs, and topics across Lattice.
This paper demonstrates image classification at multi-kHz frame rates using single-pixel imaging (SPI) with a microLED-on-CMOS digital light projector for ultrafast pattern generation. They bypass image reconstruction by using a spatiotemporal transformation of information and compare the classification accuracy on MNIST using an extreme learning machine (ELM) and a shallow backpropagation-trained neural network. The SPI-based ELM achieves promising results as a binary classifier, suggesting potential for efficient anomaly detection in ultrafast imaging.
Forget slow image sensors: this work achieves multi-kHz image classification by bypassing image reconstruction entirely with single-pixel imaging and a spatiotemporal transformation.
Pattern recognition and image classification are essential tasks in machine vision. Autonomous vehicles, for example, require being able to collect the complex information contained in a changing environment and classify it in real time. Here, we experimentally demonstrate image classification at multi-kHz frame rates combining the technique of single pixel imaging (SPI) with a low complexity machine learning model. The use of a microLED-on-CMOS digital light projector for SPI enables ultrafast pattern generation for sub-ms image encoding. We investigate the classification accuracy of our experimental system against the broadly accepted benchmarking task of the MNIST digits classification. We compare the classification performance of two machine learning models: An extreme learning machine (ELM) and a backpropagation trained deep neural network. The complexity of both models is kept low so the overhead added to the inference time is comparable to the image generation time. Crucially, our single pixel image classification approach is based on a spatiotemporal transformation of the information, entirely bypassing the need for image reconstruction. By exploring the performance of our SPI based ELM as binary classifier we demonstrate its potential for efficient anomaly detection in ultrafast imaging scenarios.
