Based on YOLOv7-Tiny Improved Model of Remote Sensing Image Detection
Yue Wang, Huaimeng Xiao, Ya Wang, Jun Huang · 2023
Object detection technology has become mature, but it is still challenging for remote sensing image detection. For example, the target size is small and difficult to separate from the surrounding background. The targets are distributed sparsely and unevenly, and the dense targets have occlusions, which makes it difficult to detect by the model. To solve this problem, this paper proposes an improved detection model based on YOLOv7-tiny. For the detection of small targets in data set, coordinate attention mechanism and loss function fusion method are used to improve the accuracy of detection of small targets. Combined with the idea of feature separation and fusion, the C5 module of YOLOv7-tiny model is improved to reduce the feature loss in the training process and improve the reasoning speed of the model. Compared with the original network model, the performance of the improved model is significantly improved, with [email protected] reaching 0.935 and FPS reaching 141.13.