Paper address : https://arxiv.org/abs/2207.02696
Source address : https://github.com/WongKinYiu/yolov7
1. Dataset
Copy the YOLOV5 dataset directly to the root directory
The Images folder contains pictures, and the Labels folder contains corresponding labels.
2. Download pre-trained weights
The corresponding model downloads the corresponding pre-training weights and places them in the root directory. This article uses yolov7.pt
3. Configure the yaml file
Create a new VOC.yaml file under the yolov7/data folder (copy COCO.yaml directly)
1. Change the data set address to the corresponding address
2. Change the number of categories and category names of the dataset
4. Modify train.py
epochs: the number of rounds of training
batch_size: YOLOV7 takes up a lot of video memory, it is recommended to set it to 2 or 4