Li Hao, Liu Changjiang, Yang Yang
IEEE Comput Graph Appl. 2023 Nov-Dec;43(6):9-21. doi: 10.1109/MCG.2023.3253167. Epub 2023 Nov 6.
We have collected a novel, nighttime scene dataset, called Rebecca, including 600 real images captured at night with pixel-level semantic annotations, which is currently scarce and can be invoked as a new benchmark. In addition, we proposed a one-step layered network, named LayerNet, to combine local features rich in appearance information in the shallow layer, global features abundant in semantic information in the deep layer, and middle-level features in between by explicitly modeling multistage features of objects in the nighttime. In addition, a multihead decoder and a well-designed hierarchical module are utilized to extract and fuse features of different depths. Numerous experiments show that our dataset can significantly improve the segmentation ability of the existing models for nighttime images. Meanwhile, our LayerNet achieves the state-of-the-art accuracy on Rebecca (65.3% mIOU). The dataset is available at https://github.com/Lihao482/REebecca.
我们收集了一个名为Rebecca的新型夜间场景数据集，其中包含600张夜间拍摄的真实图像，并带有像素级语义标注，目前此类数据集较为稀缺，可作为一个新的基准。此外，我们提出了一种一步分层网络，名为LayerNet，通过显式建模夜间物体的多阶段特征，将浅层中富含外观信息的局部特征、深层中富含语义信息的全局特征以及两者之间的中层特征相结合。此外，还利用多头解码器和精心设计的分层模块来提取和融合不同深度的特征。大量实验表明，我们的数据集可以显著提高现有模型对夜间图像的分割能力。同时，我们的LayerNet在Rebecca数据集上达到了当前最优的准确率（平均交并比为65.3%）。该数据集可在https://github.com/Lihao482/REebecca获取。
