We provide an innovative methodology for detecting small objects in remote sensing imagery. Our method addresses challenges related to missed and false detections caused by the limited pixel representation of small objects. It integrates super-resolution technology with dynamic feature fusion to enhance detection accuracy. We introduce a cross-stage local feature fusion module to improve feature extraction. In addition, we propose a super-resolution network with soft thresholding to refine small object features, resulting in improving resolution of feature maps while reducing redundancy. Furthermore, we embed a dynamic fusion module based on feature space relationships into a dual-branch network to strengthen the role of the super-resolution branch. Experimental validation on DIOR and NWPU VHR-10 datasets shows mAP improvements to 73.9% and 93.7%, respectively, with FLOPs of 24.89G and 22.33G. Our method outperforms existing approaches regarding accuracy and number of parameters, effectively addressing challenges in small object detection in remote sensing imagery.

Many previous works have achieved tremendous success for point cloud processing. However, they still suffer from inefficiency in memory and computation. In this paper, we introduce Lightweight Hierarchical Parallel Heterogeneous Group Convolutional Neural Networks (LHPHGCNN), an efficient and lightweight neural architecture to achieve better performance but lower complexity than most existing methods for point cloud processing. By designing different local structure encodings, LHPHGCNN fully mines rich local geometric features. Additionally, we further propose Hierarchical Parallel Heterogeneous Group Convolution (HPHGConv) to simultaneously capture the discriminative nonlocal features and fine-grained local geometric features of point clouds in heterogeneous groups with fewer parameters and lower computing costs, which helps to recognize elusive shapes. To further capture the contextual features along with rich semantics, we introduce a novel multi-scale semantics (MSS) strategy to progressively increase the receptive field for each local area through the information communication between different scale areas. Extensive experiments show that our LHPHGCNN significantly outperforms state-of-the-art approaches for shape classification on ModelNet40 and semantic segmentation on three large scale benchmarks S3DIS, vKITTI, ScanNet, and SemanticKITTI in terms of accuracy and complexity.