Rotation Equivariant Feature Image Pyramid Network for Object Detection in Optical Remote Sensing Imagery

Abstract

Efficient object detection methods have recently received great attention in remote sensing. Although deep convolutional networks often have excellent detection accuracy, their deployment on resource-limited edge devices is difficult. Knowledge distillation (KD) is a strategy for addressing this issue since it makes models lightweight while maintaining accuracy. However, existing KD methods for object detection have encountered two constraints. First, they discard potentially important background information and only distill nearby foreground regions. Second, they only rely on the global context, which limits the student detector's ability to acquire local information from the teacher detector. To address the aforementioned challenges, we propose Attention-based Feature Distillation (AFD), a new KD approach that distills both local and global information from the teacher detector. To enhance local distillation, we introduce a multi-instance attention mechanism that effectively distinguishes between background and foreground elements. This approach prompts the student detector to focus on the pertinent channels and pixels, as identified by the teacher detector. Local distillation lacks global information, thus attention global distillation is proposed to reconstruct the relationship between various pixels and pass it from teacher to student detector. The performance of AFD is evaluated on two public aerial image benchmarks, and the evaluation results demonstrate that AFD in object detection can attain the performance of other state-of-the-art models while being efficient.

Spatial Attention Guided Residual Attention Network for Hyperspectral Image Classification

“…To verify the effectiveness, the proposal is compared with two traditional classical methods: SVM with a radial basis function kernel and RF, and ten well discussed DL-based methods: 2-D CNN [41], 3-D CNN [41], spectral spatial residual network (SSRN) [46], spectral spatial attention network (SSAN) [62], center attention network (CAN) [63], double-branch multi-attention mechanism network (DBMA) [59], double-branch dual-attention mechanism network (DBDA) [68], 3-D cascaded spectral-spatial element attention network (CSSEAN) [67], residual spectral spatial attention network (RSSAN) [57], and rotation equivariant feature image pyramid network (REFIPN) [56]. For each method, the network from the original article is adopted.…”

“…Besides the spectral attention module, the spatial attention module was also proposed to enhance the significance of the relevant spatial regions. For example, Shamsolmoali [56] et al employed a spatial attention module to increase the discriminating ability of network during the feature fusion. By embedding the spectral attention and spatial attention modules into the residual blocks sequentially, the useful spectral-spatial features are obtained to improve the classification performances [50], [57]- [59].…”

Multiscale attention-based detection of tiny targets in aerial beach images

“…The algorithm uses upsampling and jumps connections to extract multi-scale features of different network depths during the training process, which improves the detection accuracy and speed of small target detection. Rotation equivariant feature image pyramid network (REFIPN) (Shamsolmoali et al, 2022b) improves the ability to focus on small targets in remote sensing images through scale adaptation. REFIPN uses a single detector in parallel with a lightweight image pyramid to extract features at a wide range of scales and orientations and generate regions of interest to improve the performance of small-scale object detection performance.…”

“…REFIPN uses a single detector in parallel with a lightweight image pyramid to extract features at a wide range of scales and orientations and generate regions of interest to improve the performance of small-scale object detection performance. Shamsolmoali et al (Shamsolmoali et al, 2022a) proposed a weakly supervised approach for object detection in remote sensing images and designed a contextual fine-grained model with significant attention to different objects and target parts. Liu et al (Liu et al, 2021a) proposed a high-resolution detection network for small targets, which improves the detection performance of small targets with reduced computational cost by using a shallow network for highresolution images and a deep network for low-resolution images.…”

Strategic Positioning of the Greater Bay Area of Guangdong, Hong Kong and Macao

“…Existing works have utilized deep neural networks to extract features from large-scale data to boost classification accuracy (Nie et al, 2022;Xiong et al, 2022;Liang et al, 2020;Fu et al, 2019), but obtaining distinct features from scattered and imbalanced datasets remains a challenge. Recent efforts have honed in on precise feature extraction in images, using techniques ranging from image pyramid networks with rotated convolutions (Shamsolmoali et al, 2021) to methods that blend low-level and high-level semantics , and attention-augmented feature representations (Nie et al, 2022;Liang et al, 2020;Zhang et al, 2022).…”