当前位置：网站首页>CVPR2020: Seeing Through Fog Without Seeing Fog

CVPR2020: Seeing Through Fog Without Seeing Fog

2022-08-11 06:16:00 【zhSunw】

Seeing Through Fog Without Seeing Fog

Thesis address: https://arxiv.org/pdf/1902.08913v3.pdf

Multimodal Adverse Weather Dataset

insert image description hereinsert image description here

Dataset address: https://www.uni-ulm.de/en/in/driveu/projects/dense-datasets
A total of 1.4 million frames were collected at a frame rate of 10 Hz.Every 100 frames are manually marked to balance scene type overlays.The generated annotation contains 5.5k sunny frames, 1k dense fog frames, 1k light fog frames, and 4k snow/rain frames.

Adaptive Deep Fusion (not detailed)

insert image description here

  1. Adaptive Multimodal Single-Shot Fusion
  2. Entropy-steered Fusion

Expriments

insert image description here

原网站

版权声明
本文为[zhSunw]所创，转载请带上原文链接，感谢
https://yzsam.com/2022/223/202208110514323026.html

边栏推荐

猜你喜欢

随机推荐