BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Dual-Pixel Raindrop Removal - Yizhou Li\, Tokyo Institute of Techn
 ology
DTSTART:20230727T130000Z
DTEND:20230727T140000Z
UID:TALK203497@talks.cam.ac.uk
CONTACT:Rafal Mantiuk
DESCRIPTION:Removing raindrops in images has been addressed as a significa
 nt task for various computer vision applications. In this paper\, we propo
 se the first method using a Dual-Pixel (DP) sensor to better address the r
 aindrop removal. Our key observation is that raindrops attached to a glass
  window yield noticeable disparities in DP's left-half and right-half imag
 es\, while almost no disparity exists for in-focus backgrounds. Therefore\
 , DP disparities can be utilized for robust raindrop detection. The DP dis
 parities also brings the advantage that the occluded background regions by
  raindrops are shifted between the left-half and the right-half images. Th
 erefore\, fusing the information from the left-half and the right-half ima
 ges can lead to more accurate background texture recovery. Based on the ab
 ove motivation\, we propose a DP Raindrop Removal Network (DPRRN) consisti
 ng of DP raindrop detection and DP fused raindrop removal. To efficiently 
 generate a large amount of training data\, we also propose a novel pipelin
 e to add synthetic raindrops to real-world background DP images. Experimen
 tal results on synthetic and real-world datasets demonstrate that our DPRR
 N outperforms existing state-of-the-art methods\, especially showing bette
 r robustness to real-world situations.\n\nTo join remotely: https://cl-cam
 -ac-uk.zoom.us/j/93259821276?pwd=UWVKTVpReXVoQUxXcFBiQnlKL0Zjdz09\n
LOCATION:SS03 - William Gates Building
END:VEVENT
END:VCALENDAR