Classification of optics-free images with deep neural networks
This work addresses privacy and power efficiency challenges in imaging for applications like surveillance or mobile devices, though it is incremental as it applies existing deep learning methods to a new type of data.
The researchers tackled the problem of classifying images captured without optical components by training deep neural networks, achieving 92% accuracy in binary classification and multi-class detection without reconstructing human-interpretable images.
The thinnest possible camera is achieved by removing all optics, leaving only the image sensor. We train deep neural networks to perform multi-class detection and binary classification (with accuracy of 92%) on optics-free images without the need for anthropocentric image reconstructions. Inferencing from optics-free images has the potential for enhanced privacy and power efficiency.