【Yolov8足球追踪】YOLO格式足球检测数据集制作及目标追踪

请添加图片描述
可以只是已经制作好的数据集，也可以进行制作：
数据集制作：1，2，3步
数据集下载： https://pan.baidu.com/s/1upT_aD06lzjuz-Xjze0dcw 提取码: 6eax
windows下分卷解压文件：https://blog.csdn.net/qq_41941900/article/details/142366183

1.sn-tracking数据集足球赛事追踪数据集下载：

https://github.com/SoccerNet/sn-tracking
代码下载：
pip install SoccerNet

from SoccerNet.Downloader import SoccerNetDownloader
mySoccerNetDownloader = SoccerNetDownloader(LocalDirectory="path/to/SoccerNet")
mySoccerNetDownloader.downloadDataTask(task="tracking", split=["train","test","challenge"])
mySoccerNetDownloader.downloadDataTask(task="tracking-2023", split=["train", "test", "challenge"])

2.提取sn-tracking数据集中只有足球的图片，生成YOLO格式的数据

import os
import shutil
import cv2

"从snmot数据集中挑选含有足球的数据"
base_path = r"D:\mywork\dataset\IKCEST\images\train"
dataset_image_path = "IKCEST/images/train"
dataset_labels_path = "IKCEST/labels/train"
os.makedirs(dataset_image_path, exist_ok=True)
os.makedirs(dataset_labels_path, exist_ok=True)
snmot_list = os.listdir(base_path)
print(snmot_list)
for path in snmot_list:
    new_path = os.path.join(base_path,path)
    gt = os.path.join(new_path,"gt/gt.txt")
    gameinfo = os.path.join(new_path, "gameinfo.ini")
    with open(gameinfo) as f:
        trackletID_ball = [i.split("=")[0].split("_")[-1] for i in  f.readlines() if "ball" in i]
    print(trackletID_ball)
    with open(gt) as f:
        label = f.readlines()
    for trackletID in trackletID_ball:
        label_list = []
        for i in label:
            seq = i.split(",")
            if seq[1] == trackletID:
                image_path = os.path.join(new_path, 'img1/{:06d}.jpg'.format(eval(seq[0]))).replace("\\","/")
                new_image_path = os.path.join(dataset_image_path,path+"_"+os.path.basename(image_path))
                label_path = os.path.join(dataset_labels_path,path+"_"+os.path.basename(image_path).replace(".jpg",".txt"))
                shutil.copy(image_path,new_image_path)
                x, y, w, h = [eval(mt) for mt in seq[2:6]]
                x_center = x + w / 2
                y_center = y + h / 2
                height, width, channels = cv2.imread(image_path).shape
                normalized_x_center = x_center / width
                normalized_y_center = y_center / height
                normalized_width = w / width
                normalized_height = h / height
                with open(label_path,"a") as wf:
                    wf.write(f"{0} {normalized_x_center} {normalized_y_center} {normalized_width} {normalized_height}\n")

3.将提取的只含有足球的数据集划分训练集和测试集

from glob import iglob
import os
import shutil
import random
dataset_image_path = "IKCEST/images/train"
dataset_labels_path = "IKCEST/labels/train"

dev_dataset_image_path = "IKCEST/images/dev"
dev_dataset_labels_path = "IKCEST/labels/dev"
os.makedirs(dev_dataset_image_path, exist_ok=True)
os.makedirs(dev_dataset_labels_path, exist_ok=True)

imageset = list(iglob(os.path.join(dataset_image_path,"*.jpg"),recursive=True))
print(len(imageset))
test_index = random.sample(list(range(len(imageset))),int(0.2*len(imageset)))
print(len(test_index))
for index in test_index:
    new_image = os.path.join(dev_dataset_image_path,os.path.basename(imageset[index]))
    new_label = os.path.join(dev_dataset_labels_path, os.path.basename(imageset[index]).replace(".jpg",".txt"))
    shutil.move(imageset[index],new_image)
    shutil.move(imageset[index].replace(".jpg",".txt").replace("images","labels"), new_label)


from glob import iglob
import os
base = "IKCEST"

train = ["train","dev"]
for i in train:
    base_path = os.path.join(base,"images",i)
    image_list = list(iglob(os.path.join(base_path,"*.jpg"),recursive=True))
    with open(base + "/" + i + ".txt", "w") as fm:
        for file_ in image_list:
            fm.write("./"+file_.replace("IKCEST","").replace("\\","/")+"\n")

4.制作Yolov8的配置文件及训练配置yaml

SNMOT.yaml

path: D:/mywork/IKCEST_Football  # dataset root dir
train: train.txt  # train images (relative to 'path') 118287 images
val: dev.txt  # val images (relative to 'path') 5000 images
test: dev.txt  # 20288 of 40670 images, submit to https://competitions.codalab.org/competitions/20794

names:
    0: ball

# Download script/URL (optional)
download: |
  from utils.general import download, Path

  # Download labels
  #segments = True  # segment or box labels
  #dir = Path(yaml['path'])  # dataset root dir
  #url = 'https://github.com/WongKinYiu/yolov7/releases/download/v0.1/'
  #urls = [url + ('coco2017labels-segments.zip' if segments else 'coco2017labels.zip')]  # labels
  #download(urls, dir=dir.parent)

  # Download data
  #urls = ['http://images.cocodataset.org/zips/train2017.zip',  # 19G, 118k images
  #        'http://images.cocodataset.org/zips/val2017.zip',  # 1G, 5k images
  #        'http://images.cocodataset.org/zips/test2017.zip']  # 7G, 41k images (optional)
  #download(urls, dir=dir / 'images', threads=3)

yolov8-p2.yaml

# Ultralytics YOLO 🚀, AGPL-3.0 license
# YOLOv8 object detection model with P2-P5 outputs. For Usage examples see https://docs.ultralytics.com/tasks/detect

# Parameters
nc: 80 # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov8n.yaml' will call yolov8.yaml with scale 'n'
  # [depth, width, max_channels]
  #n: [0.33, 0.25, 1024]
  #s: [0.33, 0.50, 1024]
  #m: [0.67, 0.75, 768]
  #l: [1.00, 1.00, 512]
  x: [1.00, 1.25, 512]

# YOLOv8.0 backbone
backbone:
  # [from, repeats, module, args]
  - [-1, 1, Conv, [64, 3, 2]] # 0-P1/2
  - [-1, 1, Conv, [128, 3, 2]] # 1-P2/4
  - [-1, 3, C2f, [128, True]]
  - [-1, 1, Conv, [256, 3, 2]] # 3-P3/8
  - [-1, 6, C2f, [256, True]]
  - [-1, 1, Conv, [512, 3, 2]] # 5-P4/16
  - [-1, 6, C2f, [512, True]]
  - [-1, 1, Conv, [1024, 3, 2]] # 7-P5/32
  - [-1, 3, C2f, [1024, True]]
  - [-1, 1, SPPF, [1024, 5]] # 9

# YOLOv8.0-p2 head
head:
  - [-1, 1, nn.Upsample, [None, 2, "nearest"]]
  - [[-1, 6], 1, Concat, [1]] # cat backbone P4
  - [-1, 3, C2f, [512]] # 12

  - [-1, 1, nn.Upsample, [None, 2, "nearest"]]
  - [[-1, 4], 1, Concat, [1]] # cat backbone P3
  - [-1, 3, C2f, [256]] # 15 (P3/8-small)

  - [-1, 1, nn.Upsample, [None, 2, "nearest"]]
  - [[-1, 2], 1, Concat, [1]] # cat backbone P2
  - [-1, 3, C2f, [128]] # 18 (P2/4-xsmall)

  - [-1, 1, Conv, [128, 3, 2]]
  - [[-1, 15], 1, Concat, [1]] # cat head P3
  - [-1, 3, C2f, [256]] # 21 (P3/8-small)

  - [-1, 1, Conv, [256, 3, 2]]
  - [[-1, 12], 1, Concat, [1]] # cat head P4
  - [-1, 3, C2f, [512]] # 24 (P4/16-medium)

  - [-1, 1, Conv, [512, 3, 2]]
  - [[-1, 9], 1, Concat, [1]] # cat head P5
  - [-1, 3, C2f, [1024]] # 27 (P5/32-large)

  - [[18, 21, 24, 27], 1, Detect, [nc]] # Detect(P2, P3, P4, P5)

5.训练足球检测模型

import sys
sys.path.append("ultralytics")
from ultralytics import YOLO
import os
os.system("wandb offline")
os.environ["KMP_DUPLICATE_LIB_OK"]="TRUE"
#https://blog.csdn.net/darkredrock/article/details/131046149
#model = YOLO('yolov8_config/yolov8-p2.yaml').load("yolov8x.pt")
model = YOLO('yolov8_config/yolov8-p2.yaml').load("runs/detect/train13/weights/last.pt")
#model = YOLO('yolov8_config/yolov8.yaml').load("runs/detect/train9/weights/last.pt")
results = model.train(data="yolov8_config/SNMOT.yaml", epochs=40, device='0',
                      batch=4,save=True, resume=True,amp=False,workers=0)  # 断点恢复训练模型

6.足球目标追踪

# #https://blog.csdn.net/qq_42452134/article/details/135241059
import cv2
from ultralytics import YOLO
import os
import torch
os.environ["KMP_DUPLICATE_LIB_OK"]="TRUE"
# # 加载YOLOv8模型
model = YOLO(r'runs\detect\train14\weights\best.pt').to("cuda:0")
# 打开视频文件
video_path = "a3.mp4"
cap = cv2.VideoCapture(video_path)


sz = (int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)),
        int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)))
fps = 30
fourcc = cv2.VideoWriter_fourcc('m', 'p', '4', 'v')

vout_1 = cv2.VideoWriter()
vout_1.open('a3_output_v8_p2_noaug.mp4',fourcc,fps,sz,True)


# 循环遍历视频帧
kk = 0
with open("result.txt","w") as f:
    while cap.isOpened():
        # 从视频读取一帧
        success, frame = cap.read()

        if success:
            # 在帧上运行YOLOv8追踪，持续追踪帧间的物体
            results = model.track(frame, persist=True)
            # 输出每次追踪推理结果的boxes，这些参数实际上是和模型直接predict类似的。
            if len(results[0].boxes.data) !=0:
                result_track = torch.cat((torch.zeros(results[0].boxes.cls.shape).unsqueeze(1).to("cpu"), results[0].boxes.xywh.to("cpu"), results[0].boxes.conf.unsqueeze(1).to("cpu")), dim=1)
                for i in range(result_track.shape[0]):
                    line = f"{kk+1},{int(result_track[i][0].item())}, {result_track[i][1].item()}, {result_track[i][2].item()}, {result_track[i][3].item()}, {result_track[i][4].item()}, {result_track[i][5].item()}, 1, -1, -1"
                    f.write(line+"\n")

                # 在帧上展示结果
                annotated_frame = results[0].plot()
                vout_1.write(annotated_frame)
                # 展示带注释的帧
                #cv2.imshow("YOLOv8 Tracking", annotated_frame)
                os.makedirs("runs/detect/track2",exist_ok=True)
                cv2.imwrite(f"runs/detect/track2/{kk}.png",annotated_frame)
            else:
                vout_1.write(frame)
        else:
            # 如果视频结束则退出循环
            break
        kk +=1