DAD vision: opto-electronic co-designed computer vision with division adjoint method
This work addresses miniaturization and efficiency challenges for mobile and embedded vision systems, though it appears incremental as it builds on existing optical and neural network concepts.
The authors tackled the computational and size limitations of mobile computer vision systems by proposing an ultra-thin diffractive optical element for passive optical convolution and a co-design method, achieving similar performance on CIFAR-10 classification without power consumption.
The miniaturization and mobility of computer vision systems are limited by the heavy computational burden and the size of optical lenses. Here, we propose to use a ultra-thin diffractive optical element to implement passive optical convolution. A division adjoint opto-electronic co-design method is also proposed. In our simulation experiments, the first few convolutional layers of the neural network can be replaced by optical convolution in a classification task on the CIFAR-10 dataset with no power consumption, while similar performance can be obtained.