Preliminary test yolov7

First try YOLOv7

Code ：GitHub - WongKinYiu/yolov7: Implementation of paper - YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors

Thesis download website ：https://arxiv.org/pdf/2207.02696.pdf

1、 Web version demo

Huggingface Provides a web presentation demo. Sign in Yolov7 - a Hugging Face Space by akhaliq, It can be on the left img Click the box to select the picture , choice model by yolov7（ You can also choose yolov7-e6） Click on Submit after , stay output You can see the output prediction results .

In addition to Example There's one down there horse Pictures of the , After clicking, you can also see the results in the output .

2、 Code runs

（1） install requirements.txt

 pip install -r requirements.txt

（2） Download model

stay github page , There is a model download connection , Click directly to download , Of course, you can also download all the models at the same time , Put it in detect.py Same directory .

（3） function detect.py

Directly in pycharm Run or enter in the terminal ：

python detect.py --weights yolov7.pt --conf 0.25 --img-size 640 --source inference/images/horses.jpg

Can be in runs\detect\exp3\horses.jpg See the prediction results .

（4） Train your own dataset

You can refer to the following article , I won't repeat .

【 Environment configuration 】 Hyperdetail ,YOLOV7 Train your own dataset ：

https://www.pudn.com/news/62c92c865f75f3409e909b15.html

yolov7 From environment configuration to training your own data set AlexeyAB edition

https://blog.csdn.net/weixin_41868104/article/details/125663323

doubt 1： If you don't download the model , function detect.py Generally, models can be downloaded automatically , However, it is found that git Update model error prompt , The specific method is still under study , Netizens can help analyze .