基于图像的目标检测方法及装置、电子设备及存储介质
By generating weight vectors through convolution and global pooling of image data, and combining feature compression and fusion of different modalities, the problem of decreased detection accuracy of convolutional neural networks is solved, and efficient target object detection is achieved.
CN115797697BActive Publication Date: 2026-07-17BEIJING SMARTCHIP MICROELECTRONICS TECHNOLOGY CO LTD +1
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING SMARTCHIP MICROELECTRONICS TECHNOLOGY CO LTD
- Filing Date
- 2022-12-06
- Publication Date
- 2026-07-17
AI Technical Summary
Technical Problem
The detection accuracy of existing convolutional neural networks saturates or decreases as depth increases, and the redundant feature propagation of residual networks affects the detection performance.
Method used
By convolving image data to generate convolutional feature vectors, global pooling and full connectivity are performed to generate weight vectors, output target feature vectors, remove redundant features, and feature compression and fusion are applied to image data of different modalities.
Benefits of technology
It improves detection efficiency and accuracy, especially when fusing multimodal data, by leveraging modal complementarity to enhance detection performance.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN115797697B_ABST
Abstract
本发明提供一种基于图像的目标检测方法、基于图像的目标检测装置、电子设备及非易失性计算机可读存储介质。方法包括对图像数据进行卷积以得到卷积特征向量；对所述卷积特征向量进行全局池化操作及全连接操作，以生成权重向量；根据所述卷积特征向量和所述权重向量，输出目标特征向量，并基于所述目标特征向量进行目标对象的检测。通过对图像数据卷积得到卷积特征向量，对卷积特征向量进行全局池化和全连接，从而得到卷积特征向量中的每个特征的权重，以生成权重向量。根据权重向量和卷积特征向量输出目标特征向量，去除了冗余特征，不仅提升了检测效率，而且保证了基于目标特征向量进行目标对象的检测的检测效果。
Need to check novelty before this filing date? Find Prior Art