The PyTorch version of YOLOV4 reappears from zero and the actual test of people and vehicles
![1602573968553870.png WeChat picture_20201013152306.png](https://s2.51cto.com/images/20201013/1602573968553870.png)
1、 A large number of manual drawings are given lectures (no propaganda is allowed)
Let you understand all the core ideas of YOLOV4
![1602572153790098.png WeChat picture_20201013145504.png](https://s2.51cto.com/images/20201013/1602572153790098.png)
![1602578972634821.png WeChat picture_20201013145507.png](https://s2.51cto.com/images/20201013/1602578972634821.png)
![1602572327501617.png WeChat picture_20201013145509.png](https://s2.51cto.com/images/20201013/1602572327501617.png)
2、 Build a complete YOLOV4 project code from an empty directory
1. 800 lines of code can help you understand all the cores of YOLOV4
1) YOLOV4 loss function, training core process build_target
2) YOLOV4 overall network structure, and all sub module components are written from scratch
![1602573319187711.png WeChat picture_20201013151336.png](https://s2.51cto.com/images/20201013/1602573319187711.png)
2. 1000+line by line code interpretation+Debug, fully understand the training code
1) Complete training code and inference code
2) Integration and explanation of various tools, including darknet model import tool, map and admission rate calculation tool, training process visualization tool
3、 Pedestrian and vehicle detection practice
After we have built all PyTorch codes, we will use the codes to complete YOLOV4 project practice, learn model tailoring skills and project parameter adjustment experience.
![1602573719511273.png WeChat picture_20201013152056.png](https://s2.51cto.com/images/20201013/1602573719511273.png)
Attached project screenshots:
![1602574107908692.png WeChat picture_20201013152809. png](https://s2.51cto.com/images/20201013/1602574107908692.png)
![1602574218160876.png WeChat picture_20201013152951.png](https://s2.51cto.com/images/20201013/1602574218160876.png)
![1602574522689647.png WeChat picture_20201013153439.png](https://s2.51cto.com/images/20201013/1602574522689647.png)