Recently, many studies have successfully exploited convolutional neural networks to improve the performance of object detection in remote sensing images. However, detecting tiny objects is still challenging for two main neglected problems: i) The features of tiny objects are insufficient and prone to aliasing during the multiresolution aggregation process; ii) Tiny objects are position-sensitive, resulting in poor localization capabilities. In this article, we propose a feature-and-spatial aligned network (FSANet) to alleviate these issues. FSANet is an anchor-free detector that utilizes the alignment mechanism and progressive optimization strategy to obtain more discriminative features and accurate localization results. Concretely, we first present a feature-aware alignment module ( ${\rm FA}^{{2}}\text{M}$ ) to align features with different resolutions during fusion. By learning transformation offsets, ${\rm FA}^{{2}}\text{M}$ reencodes pixel-spatial information between feature maps of adjacent levels and adaptively adjusts the regular feature interpolation. Additionally, a spatial-aware guidance head (SAGH) is introduced to iteratively optimize network predictions in a coarse-to-fine fashion. SAGH first predicts the object shape at each spatial location on feature maps. For more precise predictions, it then captures geometry-aware convolutional features accordingly to update the coarse localization estimations. Extensive experiments are conducted on three tiny object detection datasets, i.e., AI-TOD, GF1-LRSD, and TinyPerson, demonstrating the effectiveness of our approach.

Recently, deep learning-based methods have made great improvements in object detection in remote sensing images (RSIs). However, detecting tiny objects in low-resolution images is still challenging. The features of these objects are not distinguishable enough due to their tiny size and confusing backgrounds and can be easily lost as the network deepens or downsamples. To address these issues, we propose an effective Tiny Ship Detector for Low-Resolution RSIs, abbreviated as LR-TSDet, consisting of three key components: a filtered feature aggregation (FFA) module, a hierarchical-atrous spatial pyramid (HASP) module, and an IoU-Joint loss. The FFA module captures long-range dependencies by calculating the similarity matrix so as to strengthen the responses of instances. The HASP module obtains deep semantic information while maintaining the resolution of feature maps by aggregating four parallel hierarchical-atrous convolution blocks of different dilation rates. The IoU-Joint loss is proposed to alleviate the inconsistency between classification and regression tasks, and guides the network to focus on samples that have both high localization accuracy and high confidence. Furthermore, we introduce a new dataset called GF1-LRSD collected from the Gaofen–1 satellite for tiny ship detection in low-resolution RSIs. The resolution of images is 16m and the mean size of objects is about 10.9 pixels, which are much smaller than public RSI datasets. Extensive experiments on GF1-LRSD and DOTA-Ship show that our method outperforms several competitors, proving its effectiveness and generality.

In recent years, convolutional neural networks (CNNs) have made significant progress in remote sensing scene classification (RSSC) tasks. Because obtaining a large number of labeled images is time-consuming and expensive and the generalization ability of supervised models is limited, domain adaptation is widely introduced into RSSC. However, existing adaptation approaches mainly aim to align the distribution of features in a single representation space, which results in losing information and limiting the spatial range for extracting domain-invariant features. In addition, some of the methods simultaneously align pixel-level (local) and image-level (global) features for better results but suffer from searching for the best weight of the two parts manually, which is time-consuming and computing-expensive. To overcome the above issues, a novel feature fusion-and-alignment approach named Multi-Representation Dynamic Adaption Network (MRDAN) is proposed for cross-domain RSSC. Concretely, a Feature-Fusion Adaptation (FFA) module is embedded into the network, which maps samples to multiple representations and fuses them to obtain a broader domain-invariant feature space. Based on this hybrid space, we introduce a cross-domain Dynamic Feature-Alignment Mechanism (DFAM) to quantitatively evaluate and adjust the relative importance of the local and global adaptation losses during domain adaptation. The experimental results on the 12 transfer tasks between the UC Merced land-use, WHU-RS19, AID, and RSSCN7 data sets demonstrate the effectiveness of the proposed MRDAN over the state-of-the-art domain adaptation methods in RSSC.