Abstract:The significant differences in target scales of remote sensing images lead to remarkable variations in visual features, posing significant challenges for feature extraction, fusion, regression, and classification. For example, models frequently struggle to capture features of targets across all scales, inadequately consider the weights and importance of features at different scales during fusion, and encounter accuracy limitations when detecting targets of varying scales. To tackle these challenges, we proposes a Scale-Robust Feature Aggregation and Diffusion Network (SRFAD-Net) for remote sensing target detection. This model includes a Scale-Robust Feature Network (SRFN), an Adaptive Feature Aggregation and Diffusion (AFAD) module, and a Focaler-GIoU Loss. SRFN extracts scale-robust features by constructing a multi-scale pyramid. It includes a downsampling (ADown) module that combines the advantages of average pooling and max pooling, effectively preserving background information and salient features. This further enhances the network's ability to handle targets of varying scales and shapes. The introduced Deformable Attention(DAttention) mechanism captures target features effectively through adaptive adjustment of the receptive field's shape and size, reducing background clutter and substantially enhancing the model's performance in detecting distant objects. In the feature fusion stage, we propose the AFAD module, which utilizes a dimension-adaptive perceptual selection mechanism and parallel depthwise convolutional operations to precisely aggregate multi-channel information. It then employs a diffusion mechanism to spread contextual information across various scales, greatly improving the network's ability to extract and fuse features across multiple scales. For the detection head, we adopt the Focaler-GIoU Loss, leveraging its advantages in handling non-overlapping bounding boxes, effectively alleviating the difficulty of localization caused by scale variations. We have undertaken experiments on two widely utilized aerial target datasets: the Remote Sensing Scene Object Detection Dataset (RSOD) and NWPU VHR-10, which is a high-resolution object detection dataset from Northwestern Polytechnical University. The findings of these experiments clearly illustrate that SRFAD-Net surpasses the performances of mainstream detectors.

SRC: transferring scale-independent features to support multi-scale object recognition with deep convolutional neural network.

Dynamic Convolution Covariance Network Using Multi-Scale Feature Fusion for Remote Sensing Scene Image Classification

SSN: Scale Selection Network for Multi-Scale Object Detection in Remote Sensing Images

Scale-Balanced Real-Time Object Detection with Varying Input-Image Resolution

A Deep CNN-Based Detection Method for Multi-Scale Fine-Grained Objects in Remote Sensing Images

SAN: Learning Relationship between Convolutional Features for Multi-Scale Object Detection

ORCNN-X: Attention-Driven Multiscale Network for Detecting Small Objects in Complex Aerial Scenes

Scene recognition with CNNs: objects, scales and dataset bias

Attention To The Scale : Deep Multi-Scale Salient Object Detection

SFSANet: Multiscale Object Detection in Remote Sensing Image Based on Semantic Fusion and Scale Adaptability

Image Recognition Using Scale Recurrent Neural Networks

Structured Object-Level Relational Reasoning CNN-Based Target Detection Algorithm in a Remote Sensing Image

Efficient Scale Divide and Conquer Network for Object Detection

A Multi-Scale Approach For Remote Sensing Scene Classification Based On Feature Maps Selection And Region Representation

Hierarchical Region Based Convolution Neural Network for Multiscale Object Detection in Remote Sensing Images

Cross-Scale Feature Fusion for Object Detection in Optical Remote Sensing Images

SRFAD-Net: Scale-Robust Feature Aggregation and Diffusion Network for Object Detection in Remote Sensing Images

Delving into the Scale Variance Problem in Object Detection

Exploiting Cross-scale Consistency for Object Detection in Aerial Images

An improved object detection algorithm based on multi-scaled and deformable convolutional neural networks

Scale-Transferrable Object Detection