Reference github repository for the paper "Defocus Deblurring Using Dual-Pixel Data". We introduce a deep neural network (DNN) architecture that uses the dual-pixel (DP) sub-aperture views to reduce defocus blur.

Reference github repository for the paper "Learning to Reduce Defocus Blur by Realistically Modeling Dual-Pixel Data". We propose a procedure to generate realistic DP data synthetically. Our synthesis approach mimics the optical image formation found on DP sensors and can be applied to virtual scenes rendered with standard computer software. Lev…

Reference github repository for the paper "Improving Single-Image Defocus Deblurring: How Dual-Pixel Images Help Through Multi-Task Learning". We propose a single-image deblurring network that incorporates the two sub-aperture views into a multitask framework. Specifically, we show that jointly learning to predict the two DP views from a single …

Facial Depth and Normal Estimation using Dual-Pixel Camera (ECCV 22)

Code & Dataset release for the ICCP'23 paper: Learning to Synthesize Photorealistic Dual-pixel Images from RGBD frames