YOLOv7 algorithm reconstruction: personal self-improvement, as simple as YOLOv5, only 31 layers

Introduction to the Refactoring Project

YOLOv7 is an excellent detection algorithm. The author wrote all the components into the yaml file to enhance the transformability, but readers who are familiar with the network structure of YOLO'v5, it is still relatively difficult to watch YOLOv7. Since I am changing the YOLOv7 model recently, it is still not as convenient as YOLOv5, so I reconstruct the YOLOv7 network according to my own ideas.

1. The amount of parameters and calculations are the same before and after reconstruction. 2. The performance is exactly the same for 105 layers before reconstruction and 31 layers after reconstruction.

Before refactoring:

After refactoring:

Refactored core code

#  重构YOLOv7-----------------------------------------------------------------------------
<