Image processing method and device, storage medium and electronic device
By introducing a target focus structure into the YOLOv5 model, the data processing steps are simplified, the problems of high computational cost and excessive memory consumption are solved, and the detection efficiency and platform applicability are improved.
CN115937606BActive Publication Date: 2026-06-19ZHEJIANG DAHUA TECH CO LTD
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHEJIANG DAHUA TECH CO LTD
- Filing Date
- 2022-12-29
- Publication Date
- 2026-06-19
Smart Images
Figure CN115937606B_ABST
Abstract
This invention provides an image processing method, apparatus, storage medium, and electronic device. The method includes: acquiring a target image; inputting the target image into a target YOLOv5 model and outputting detection results for the target image, wherein the target YOLOv5 model includes at least a target focus structure, the target focus structure including a first slice layer composed of one slice layer, a first stitching layer composed of two concat layers, a second slice layer composed of two slice layers, a second stitching layer composed of four concat layers, and a third stitching layer composed of one concat layer; and returning the detection results to a client. This invention solves the problem in related technologies where the focus structure in the YOLOv5 model is complex and computationally intensive, resulting in low detection efficiency of the YOLOv5 model for target objects in images.
Need to check novelty before this filing date? Find Prior Art