• Technical Name
  • Dynamic Video Segmentation Network
  • Operator
  • National Tsing Hua University
  • Booth
  • Online display only
  • Contact
  • 李濬屹
  • Email
  • cylee@cs.nthu.edu.tw
Technical Description \"We present a detailed design of dynamicvideo segmentation network (DVSNet) for fastefficientsemantic video segmentation. DVSNet consists of two convolutional neural networks: a segmentation network anda flow network. The former generates highly accurate se-mantic segmentations, but is deeperslower. The latteris much faster than the former, but its output requires further processing
Scientific Breakthrough \"The contributions of this work are as follows:1. A frame division technique to apply different segmentationstrategies to different frame regions for maximizingthe usage of video redundancycontinuity.2. A DN for determining whether to assign an input frameregion to the segmentation network,adaptively adjustingthe update period of the key frames.3. An adaptive key frame scheduling
Industrial Applicability \"The main applications focused in this project are visual recognitionapplications for smart robot. W focus on developing deep learningtechniques for video semantic segmentation. The techniquesdeveloped in this project can be used not only for robotic industry but also for video surveillance industry.\"
本網站使用您的Cookie於優化網站。繼續瀏覽網站即表示您同意本公司隱私權政策,您可至隱私權政策了解詳細資訊。