Raindrop Clarity: A Dual-Focused Dataset for Day and Night Raindrop Removal

TL;DR

Raindrop Clarity introduces a large-scale real-world benchmark that explicitly covers both raindrop-focused and background-focused imagery under day and night conditions, addressing the gap left by prior datasets that predominantly feature daytime, background-focused raindrops. The dataset comprises 6,742 pairs and 8,444 triplets (15,186 high-quality items) across daytime and nighttime, collected with stationary capture setups and diverse devices to ensure varied raindrop shapes, sizes, and occlusions. The authors benchmark multiple raindrop removal methods and restoration backbones on daytime and nighttime splits, showing that current state-of-the-art approaches struggle with raindrop-focused and nighttime scenarios, thereby highlighting open challenges. By providing rich paired and triplet data, the work aims to推动 the development of robust raindrop removal and background restoration techniques applicable across day-night conditions and raindrop focus, with public release to catalyze progress.

Abstract

Existing raindrop removal datasets have two shortcomings. First, they consist of images captured by cameras with a focus on the background, leading to the presence of blurry raindrops. To our knowledge, none of these datasets include images where the focus is specifically on raindrops, which results in a blurry background. Second, these datasets predominantly consist of daytime images, thereby lacking nighttime raindrop scenarios. Consequently, algorithms trained on these datasets may struggle to perform effectively in raindrop-focused or nighttime scenarios. The absence of datasets specifically designed for raindrop-focused and nighttime raindrops constrains research in this area. In this paper, we introduce a large-scale, real-world raindrop removal dataset called Raindrop Clarity. Raindrop Clarity comprises 15,186 high-quality pairs/triplets (raindrops, blur, and background) of images with raindrops and the corresponding clear background images. There are 5,442 daytime raindrop images and 9,744 nighttime raindrop images. Specifically, the 5,442 daytime images include 3,606 raindrop- and 1,836 background-focused images. While the 9,744 nighttime images contain 4,838 raindrop- and 4,906 background-focused images. Our dataset will enable the community to explore background-focused and raindrop-focused images, including challenges unique to daytime and nighttime conditions. Our data and code are available at: \url{https://github.com/jinyeying/RaindropClarity}

Figures (13)

• Figure 1: Existing raindrop datasets (e.g, qian2018attentivequan2021removingporav2019cansoboleva2021raindrops) exhibit two limitations: first, they do not include raindrop-focused images (Fig. \ref{['fig:dataset']}right); and second, they lack night raindrops (Fig. \ref{['fig:dataset']}bottom).
• Figure 2: Motivation for Raindrop Clarity: real images sourced from the Internet feature scenarios overlooked by existing datasets, including raindrop-focused (right) and nighttime raindrops (bottom). In this figure, Rd = Raindrop, Bg = Background.
• Figure 3: Raindrop Clarity: Examples of our pairs $(\mathop{\mathrm{\Tilde{\mathbf{x}}}}\nolimits,{\boldsymbol b}_0)$ and triplets $(\mathop{\mathrm{\Tilde{\mathbf{x}}}}\nolimits,{\boldsymbol x}_0,{\boldsymbol b}_0)$ of a raindrop image $\mathop{\mathrm{\Tilde{\mathbf{x}}}}\nolimits$, the corresponding blurry background image ${\boldsymbol x}_0$ and the corresponding clear background image ${\boldsymbol b}_0$, in day and night. For the pairs $(\mathop{\mathrm{\Tilde{\mathbf{x}}}}\nolimits,{\boldsymbol b}_0)$, there is no ${\boldsymbol x}_0$, since in this case ${\boldsymbol x}_0 = {\boldsymbol b}_0$. The reason we have the blurry background image ${\boldsymbol x}_0$ in triplets is to have the raindrop difference maps $\mathbf{\tilde{m}}$ (see Fig. \ref{['fig:mask']}).
• Figure 4: Pairs $(\mathop{\mathrm{\Tilde{\mathbf{x}}}}\nolimits,{\boldsymbol b}_0)$ show various sizes, shapes, and densities.
• Figure 5: The triplets $(\mathop{\mathrm{\Tilde{\mathbf{x}}}}\nolimits,{\boldsymbol x}_0,{\boldsymbol b}_0)$ cover diverse scenes and shapes.