Recently, DUT-RU ISE Geometric Computing and Intelligent Media Technology Team has harvested innovative achievements at a leading conference in the field of computer vision. A paper was accepted by the CVPR 2023 conference, and a team won the championship in the CVPR 2023 Workshop of Mobile Intelligent Photography and Imaging (MIPI).

The paper “Pixels, Regions, and Objects: Multiple Enhancement for Salient Object Detection”, co-authored by Associate Professor Wang Yi, Professor Fan Xin from our school, Ruili Wang from Massey University, New Zealand, and He Xiangjian from the University of Nottingham Ningbo China, proposed a novel Multiple Enhancement Network (MENet) to simulate humans’ ability to repeatedly and gradually enhance the cognition of complex objects from the perspectives of pixels, regions and objects of the image. It also solved the problem that most existing salient object detection (SOD) had explained images from the perspective of pixels with a simple end-to-end training method, leading to the incomplete and blurred boundaries of objects in complex scenes.

PhD candidate Liu Yuqing and Associate Professor Jia Qi from our school, together with the Institute of Automation, Chinese Academy of Sciences and Xiaomi Corporation, participated in the CVPR 2023 Workshop of Mobile Intelligent Photography and Imaging (MIPI) and won the championship in the two tracks of “RGBW data joint remosaic and denoising” and “RGBW data joint fusion and denoising”.

Figure 1 RGBW data joint remosaic and denoising scheme

Figure 2 RGBW joint fusion and denoising scheme

It is known that CVPR is the top conference in the field of computer vision. On the list of 2022 Scholar Metrics released by Google, CVPR ranks the 4th (after Nature, NEJM, and Science), with the H5-index of 389, and ranks the 1st among all journals or conferences in computer science. This year’s CVPR received a total number of 9,155 full submissions, and accepted 2,360 papers with an acceptance rate of 25.78 percent.