Images captured in rainy days suffer from noticeable degradation of scene visibility.Unmanned aerial vehicles(UAVs),as important outdoor image acquisition systems,demand a proper rain removal algorithm to improve visu...Images captured in rainy days suffer from noticeable degradation of scene visibility.Unmanned aerial vehicles(UAVs),as important outdoor image acquisition systems,demand a proper rain removal algorithm to improve visual perception quality of captured images as well as the performance of many subsequent computer vision applications.To deal with rain streaks of different sizes and directions,this paper proposes to employ convolutional kernels of different sizes in a multi-path structure.Split attention is leveraged to enable communication across multiscale paths at feature level,which allows adaptive receptive field to tackle complex situations.We incorporate the multi-path convolution and the split attention operation into the basic residual block without increasing the channels of feature maps.Moreover,every block in our network is unfolded four times to compress the network volume without sacrificing the deraining performance.The performance on various benchmark datasets demonstrates that our method outperforms state-of-the-art deraining algorithms in both numerical and qualitative comparisons.展开更多
Geodashboards are often designed with explanatory elements guiding users.These elements(e.g.legends or annotations)need to be carefully designed to mitigate split attention or information integration issues.In this pa...Geodashboards are often designed with explanatory elements guiding users.These elements(e.g.legends or annotations)need to be carefully designed to mitigate split attention or information integration issues.In this paper,we report expert interviews followed by a controlled experiment where we compare two interface designs with a focus on the split attention effect:(1)a multiple-legend layout with explanatory elements located next to each view,and(2)a single-legend layout with all explanatory elements gathered in one place.Different legend layouts did not affect the performance,but affected user satisfaction.75%of the participants preferred the multiple-legend layout,and rated it with a higher usability score,mainly attributing this preference to the proximity of legend elements to the view of interest.Eye tracking data strongly and clearly verifies that participants indeed make use of the proximity:With the single-legend,the majority of eye-movement transitions were between the single-legend and the closest view to the legend,whereas with multiple-legend participants have shorter and more frequent legend visits,as well as more transitions between legends and views.Taken together,the design lesson we learned from this experiment can be summarized as‘split the legend elements,but make it close to the explained elements’.展开更多
基金the Fundation of Graduate Innovation Center in Nanjing University of Aeronautics and Astronautics(No.kfjj20191601).
文摘Images captured in rainy days suffer from noticeable degradation of scene visibility.Unmanned aerial vehicles(UAVs),as important outdoor image acquisition systems,demand a proper rain removal algorithm to improve visual perception quality of captured images as well as the performance of many subsequent computer vision applications.To deal with rain streaks of different sizes and directions,this paper proposes to employ convolutional kernels of different sizes in a multi-path structure.Split attention is leveraged to enable communication across multiscale paths at feature level,which allows adaptive receptive field to tackle complex situations.We incorporate the multi-path convolution and the split attention operation into the basic residual block without increasing the channels of feature maps.Moreover,every block in our network is unfolded four times to compress the network volume without sacrificing the deraining performance.The performance on various benchmark datasets demonstrates that our method outperforms state-of-the-art deraining algorithms in both numerical and qualitative comparisons.
基金supported by the National Science Centre,Poland[grant number UMO-2018/31/D/HS6/02770]the Norwegian Research Council[grant number 235490].
文摘Geodashboards are often designed with explanatory elements guiding users.These elements(e.g.legends or annotations)need to be carefully designed to mitigate split attention or information integration issues.In this paper,we report expert interviews followed by a controlled experiment where we compare two interface designs with a focus on the split attention effect:(1)a multiple-legend layout with explanatory elements located next to each view,and(2)a single-legend layout with all explanatory elements gathered in one place.Different legend layouts did not affect the performance,but affected user satisfaction.75%of the participants preferred the multiple-legend layout,and rated it with a higher usability score,mainly attributing this preference to the proximity of legend elements to the view of interest.Eye tracking data strongly and clearly verifies that participants indeed make use of the proximity:With the single-legend,the majority of eye-movement transitions were between the single-legend and the closest view to the legend,whereas with multiple-legend participants have shorter and more frequent legend visits,as well as more transitions between legends and views.Taken together,the design lesson we learned from this experiment can be summarized as‘split the legend elements,but make it close to the explained elements’.
