Owing to the rapid development of deep neural networks, prominent advances have been recently achieved in the semantic segmentation of remote sensing images. As the vital components of computer vision, semantic segmentation, and edge detection have strong correlation whether in the extracted features or task objective. Prior studies treated edge detection as a postprocessing operation to semantic segmentation, or they implicitly combined the two tasks. We consider that pixels around the edges are easy to be misdivided because of the prevalence of intraclass inconsistencies and interclass indistinctions, which reflect the discriminative ability of models to distinguish different classes. In this letter, we propose a multipath atrous module to first enrich the deep semantic information. Then, we combine the enhanced deep semantic information and dilated edge information generated by canny and morphological operations to obtain edge-region maps via edge-region detection module, which identifies pixels around the edges. Then, we relearn these error-prone pixels using a guidance module for the segmentation branch in a progressive guided manner. Combined with edge and segmentation branches, our progressive edge guidance network achieves an overall accuracy of 91.0% on the ISPRS Vaihingen test set, which is the new state-of-the-art result.

Global context information is vital in visual understanding problems, especially in pixel-level semantic segmentation. The mainstream methods adopt the self-attention mechanism to model global context information. However, pixels belonging to different classes usually have weak feature correlation. Modeling the global pixel-level correlation matrix indiscriminately is extremely redundant in the self-attention mechanism. In order to solve the above problem, we propose a hierarchical context network to differentially model homogeneous pixels with strong correlations and heterogeneous pixels with weak correlations. Specifically, we first propose a multi-scale guided pre-segmentation module to divide the entire feature map into different classed-based homogeneous regions. Within each homogeneous region, we design the pixel context module to capture pixel-level correlations. Subsequently, different from the self-attention mechanism that still models weak heterogeneous correlations in a dense pixel-level manner, the region context module is proposed to model sparse region-level dependencies using a unified representation of each region. Through aggregating fine-grained pixel context features and coarse-grained region context features, our proposed network can not only hierarchically model global context information but also harvest multi-granularity representations to more robustly identify multi-scale objects. We evaluate our approach on Cityscapes and the ISPRS Vaihingen dataset. Without Bells or Whistles, our approach realizes a mean IoU of 82.8% and overall accuracy of 91.4% on Cityscapes and ISPRS Vaihingen test set, achieving state-of-the-art results.

This report introduces the technical details of the team FuXi-Fresher for LVIS Challenge 2021. Our method focuses on the problem in following two aspects: the long-tail distribution and the segmentation quality of mask and boundary. Based on the advanced HTC instance segmentation algorithm, we connect transformer backbone(Swin-L) through composite connections inspired by CBNetv2 to enhance the baseline results. To alleviate the problem of long-tail distribution, we design a Distribution Balanced method which includes dataset balanced and loss function balaced modules. Further, we use a Mask and Boundary Refinement method composed with mask scoring and refine-mask algorithms to improve the segmentation quality. In addition, we are pleasantly surprised to find that early stopping combined with EMA method can achieve a great improvement. Finally, by using multi-scale testing and increasing the upper limit of the number of objects detected per image, we achieved more than 45.4% boundary AP on the val set of LVIS Challenge 2021. On the test data of LVIS Challenge 2021, we rank 1st and achieve 48.1% AP. Notably, our APr 47.5% is very closed to the APf 48.0%.

Semantic segmentation in high resolution aerial image is faced with a challenge caused by ubiquitous fine-structure objects. Traditional encoder-decoder structure losses some detail information during the process of down-sampling, which is harmful to the location of fine-structure objects. In this work, we present a multi-resolution dense encoder and stack decoder network to deal with this problem. On the one hand, the dense encoder embeds shallow detailed feature into deep semantic feature through proposed information-reserved down-sampling method called CE-Pooling. On the other hand, the stack decoder gradually enhances the detailed feature through iterative attention fusion. Extensive experiments on several benchmark datasets have been conducted, which shows that our method is superior than the state-of-the-art approaches.