YOLOv8 algorithm improvement [NO.87] Introducing the Light-weight Context Guided improvement C2_f of the Context Guided Network (CGNet)

 Foreword The YOLO algorithm improvement series is here. Many friends ask how to choose the best improvement. Below I will publish articles and guide articles based on my many years of writing. From experience, sort according to the order of priority to explain the order selection of YOLO algorithm improvement methods. Students who have specific needs can communicate with me privately:
     

First, innovate the backbone feature extraction network and improve the entire Backbone into other networks, such as the entire method in this article, Replace Backbone directly.The reason is that if this improvement is effective, then the improvement points are worth writing about. It is not just a pile of blocks, it can also be said to be a new algorithm, so If you are doing experiments, it is recommended that friends give priority to trying this modification method.

Second,Innovate feature fusion network, which is the same as the first, such as improving the original yolo algorithm PANet structure to Bifpn, etc.

Third,Improve the backbone feature extraction network, which is similar to adding an attention mechanism. Based on personal experimental conditions