Construction and practice of yolov7 in hands-on teaching

Pay attention to the parallel stars

Never get lost

Institute of computer vision

official account ID｜ComputerVisionGzq

Study Group ｜ Scan the code to get the join mode on the homepage

Computer Vision Institute column

author ：Edison_G

yolo Series has become a common target detection network in your study and work , Sure enough ,yolov7 Here we go again . It is said that the author is yolov4 Of , Let's practice .

01

summary

For details, please refer to CSDN Of he_eeeeeeeeeee, You can enter the home page or a more detailed operation process .

Source code ：https://github.com/WongKinYiu/yolov7
The paper ：https://arxiv.org/abs/2207.02696

Yolov7： The latest and fastest real-time detection framework , The most detailed analysis and explanation （ With source code ）

02

experiment

Set up the environment , It's directly used here conda According to the source code requirements.txt Just install it .

• Data preparation

Off the shelf data （VisDrone）+ The configuration script (yolov7.yaml etc. )：

Data sets ： Official VisDrone Format and yolov5 Different , I want to change it

Make your own dataset （lableme Mark your picture +labelme turn yolo Script for ）：

It is strongly recommended to use labelme

labelme After the format is marked, one figure corresponds to one json：

yolo The format is that one graph corresponds to one txt：

I transformed it txt and json Under the same folder ：

• Labeling software ：

labelme Just one exe file 50m about , Download address ：8q3m,windows Directly execute after uploading and downloading , Very convenient

• labelme turn yolov5 Script ：

（ Only the detected ones are written ,labelme Segmentation 、 Key points, etc. the script has the opportunity to write a general ）

import os
import json
import glob
# Input port , It's your pictures and json The file stored , Output txt Also in this folder 
labelme_dir=r"C:\Users\ttao\Pictures\my_"




def get_labelme_data(labelme_dir):


    with open(labelme_dir) as f:
        j=json.load(f)
        out_data=[]
        img_h =j["imageHeight"]
        img_w =j["imageWidth"]
        for shape in j["shapes"]:
            label=shape["label"]
            points=shape["points"]
            x,y,x2,y2=points[0][0],points[0][1],points[1][0],points[1][1]
            x_c=(x+x2)//2
            y_c=(y+y2)//2
            w=abs(x-x2)
            h=abs(y-y2)
            out_data.append([label,x_c,y_c,w,h])
    return img_h,img_w,out_data


def rename_Suffix(in_,mode=".txt"):
    in_=in_.split('.')
    return  in_[0]+mode


def make_yolo_data(in_dir):
    json_list=glob.glob(os.path.join(in_dir,'*.json'))


    for json_ in json_list:
        json_path=os.path.join(in_dir,json_)
        json_txt=rename_Suffix(json_)
        img_h,img_w,labelme_datas=get_labelme_data(json_path)
        with open(os.path.join(in_dir,json_txt),'w+') as f:
            for labelme_data in labelme_datas:
                label=labelme_data[0]
                x_c=labelme_data[1]/img_w
                y_c=labelme_data[2]/img_h
                w=labelme_data[3]/img_w
                h=labelme_data[4]/img_h
                f.write("{} {} {} {} {}\n".format(label,x_c,y_c,w,h))
            f.close()




if __name__ == '__main__':
    make_yolo_data(labelme_dir)

• Training

To configure ：train.py+ Dataset configuration file （.yaml）+ Select the configuration file of the network （ The default is yolov7.yaml）
Visdrone Data sets
The data set is taken from the angle of UAV , Chinese cities , Mainly all kinds of cars 、 Human data set
My specific configuration is as follows ：

train.py

–cfg  Model selection , The default here is yolov7
–data  Data format configuration , I have my own VisDrone data format , Training yolo It needs to be changed , Look at the training data set below 
–hyp  Hyperparameters , I don't care 
–batch-size  This yolov7 I set it high and easy to explode , Smaller 

VisDrone.yaml

Then you can train

 THE END 

Please contact the official account for authorization.

The learning group of computer vision research institute is waiting for you to join ！

ABOUT

Institute of computer vision

The Institute of computer vision is mainly involved in the field of deep learning , Mainly devoted to face detection 、 Face recognition , Multi target detection 、 Target tracking 、 Image segmentation and other research directions . The Research Institute will continue to share the latest paper algorithm new framework , The difference of our reform this time is , We need to focus on ” Research “. After that, we will share the practice process for the corresponding fields , Let us really experience the real scene of getting rid of the theory , Develop the habit of hands-on programming and brain thinking ！

VX：2311123606