如何訓(xùn)練帶旋轉(zhuǎn)參數(shù)的 YOLOV5 模型

代碼倉庫

git clone https://github.com/hukaixuan19970627/yolov5_obb.git

數(shù)據(jù)集準(zhǔn)備

數(shù)據(jù)標(biāo)注

rolabelimg 安裝

參考： rolabelimg

windows 下虛擬環(huán)境進(jìn)行操作

git clone https://github.com/cgvict/roLabelImg.git
cd rolabelimg
pyrcc5 -o resources.py resources.qrc 

** 運(yùn)行 **

python rolabelimg.py

標(biāo)注快捷鍵

 w : 創(chuàng)建矩形框
 e : 創(chuàng)建帶旋轉(zhuǎn)的矩形框
   z篓吁，x, c, v : 【大/小】幅度進(jìn)行【順時(shí)針/逆時(shí)針】旋轉(zhuǎn)
 a : 上一張
 d : 下一張
 ctrl + s : 保存（重要）

數(shù)據(jù)預(yù)處理

import os
import xml.etree.ElementTree as ET
import cv2
import random
from tqdm import tqdm
from multiprocessing import Pool
import numpy as np
random.seed(0)
import shutil
import math

pi = math.pi
img_path = r"./ADD/albums/images/"
anno_path = r"./ADD/albums/Annotations/"

label_path = r"./ADD/albums/labelTxt/"
# with open("jsxs_data.txt","w") as F:

classes = ['daozha']
# classes =  ['sly_dmyw', 'yw_gkxfw', 'yw_nc', 'gbps', 'xmbhyc', 'xmbhzc', 'kgg_ybh', 'kgg_ybf', 'wcaqm', 'aqmzc', 'wcgz', 'gzzc', 
# 'xy', 'jyz_pl','byq_hxq', 'hxq_gjbs', 'hxq_gjzc', 'hxq_gjtps', 'ywzt_yfyc', 'bj_bpmh', 'bj_bpps', 'bj_wkps', 'bjdsyc', 'bjzc', 'bj',
# 'bj_sxb','sxb_bjdsyc','sxb_bjdszc']

# with open("jsxs_data.txt","w") as F:
root_path = './'
ftrain = open(root_path+'trainb.txt', 'w')
ftest = open(root_path+'testb.txt', 'w')
train_percent = 0.8


files = os.listdir(anno_path)
num = len(files)
print('num image',num)
list = range(num)
tr = int(num * train_percent)
train_list = random.sample(list, tr)
print('len train',train_list)
if not os.path.exists(label_path):
    os.makedirs(label_path)

def resi(num):
    x = round(num, 6)
    x = str(abs(x))
    while len(x) < 8:
        x = x + str(0)
    return x

def convert(size, box):
    dw = 1./size[0]
    dh = 1./size[1]
    x = (box[0] + box[1])/2.0 # x = x軸中點(diǎn)
    y = (box[2] + box[3])/2.0 # y = y軸中點(diǎn)
    w = box[1] - box[0] #w = width
    h = box[3] - box[2] # h = height
    x = resi(x*dw)
    w = resi(w*dw)
    y = resi(y*dh)
    h = resi(h*dh)
    return (x,y,w,h)

def convert2(cx,cy,w,h,a,xmax,ymax):
    if a>=pi:
        a-=pi
    #計(jì)算斜徑半長(zhǎng)
    l=math.sqrt(w**2+h**2)/2
    #計(jì)算初始矩形角度
    a0=math.atan(h/w)
    #旋轉(zhuǎn),計(jì)算旋轉(zhuǎn)角
    #右上角點(diǎn) ↗
    a1=a0+a
    x1=int(cx+l*math.cos(a1)) if int(cx+l*math.cos(a1))<xmax else xmax
    y1=int(cy+l*math.sin(a1)) if int(cy+l*math.sin(a1))<ymax else ymax
    #右下角點(diǎn) ↘
    a2=-a0+a
    x2=int(cx+l*math.cos(a2)) if int(cx+l*math.cos(a2))<xmax else xmax
    y2=int(cy+l*math.sin(a2)) if int(cy+l*math.sin(a2))<ymax else ymax
    #左下角點(diǎn) ↙
    a3=a1+pi
    x3=int(cx+l*math.cos(a3)) if int(cx+l*math.cos(a3))<xmax else xmax
    y3=int(cy+l*math.sin(a3)) if int(cy+l*math.sin(a3))<ymax else ymax
    #左上角點(diǎn) ↖
    a4=a2+pi
    x4=int(cx+l*math.cos(a4)) if int(cx+l*math.cos(a4))<xmax else xmax
    y4=int(cy+l*math.sin(a4)) if int(cy+l*math.sin(a4))<ymax else ymax
    return [x1,y1,x2,y2,x3,y3,x4,y4,classes[0],0]

import glob
def process(anno_path,name):
    global train_list
    
    found_flag = 0
    img_names = ['.jpg','.JPG','.PNG','.png']
    for j in img_names:
        img_name = os.path.splitext(name)[0] + j
        if os.path.exists(img_path + img_name):
            break
    
    xml_name = os.path.splitext(name)[0] + ".xml"
    txt_name = os.path.splitext(name)[0] + ".txt"
    string1 = ""
    # print(name)
    w,h = None, None
    xml_file = ET.parse(anno_path + xml_name)
    
    root = xml_file.getroot()
    

    try:
        with open(img_path + img_name, 'rb') as f:
            check = f.read()[-2:]
        if check != b'\xff\xd9':
            print('JPEG File collapse:', img_path + img_name)
            a = cv2.imdecode(np.fromfile(img_path + img_name,dtype=np.uint8),-1)
            cv2.imencode(".jpg", a)[1].tofile(img_path + img_name)
            height,width = cv2.imdecode(np.fromfile(img_path + 
                                                   img_name,dtype=np.uint8),-1).shape[:2]
            print('----------Rewrite & Read image successfully----------')

        else:
            height,width = cv2.imdecode(np.fromfile(img_path + img_name,dtype=np.uint8),-1).shape[:2]
    except:
        print(img_path + img_name)

    if (width is not None) and (height is not None):
        count = 0
        for child in root.findall('object'):
            if child != '':
                count = count + 1
        if count != 0:

            string1 = []
            for obj in root.iter('object'):
                cls = obj.find('name').text
                
                if cls in classes:
                    cls_id = classes.index(cls)
                else:
                    print(cls)
                    continue
                xmlbox = obj.find('robndbox')
                b = (float(xmlbox.find('cx').text), float(xmlbox.find('cy').text), float(xmlbox.find('w').text),
                     float(xmlbox.find('h').text),float(xmlbox.find('angle').text))
                cx,cy,w,h,a = b
                bb = convert2(cx,cy,w,h,a,width,height) 
                #[x1,y1,x2,y2,x3,y3,x4,y4,"daozha",0]
                # for a in bb:
                #     if float(a) > 1.0:
                #         print(anno_path + xml_name + "wrong xywh",bb)
                #         return

                string1.append( " ".join([str(a) for a in bb]) + '\n') 
            out_file = open(label_path + txt_name, "w")
            for string in string1:
                out_file.write(string)
            out_file.close()
        else:
            print('count=0')
            print(img_name,"write no label txt file")
            out_file = open(label_path + txt_name, "w")
        
    else:
        print('wh is none')


def main():
    
    pbar = tqdm(total=(len(files)))
    update = lambda *args: pbar.update()
    pool = Pool(6)

    for i, name in enumerate(files):
        pool.apply_async(process, args=(anno_path,name), callback=update)
        # pbar.update(1)
    
    pool.close()
    pool.join()
    img_names = ['.jpg','.JPG','.PNG','.png']
    for i, name in enumerate(files):
        for j in img_names:
            img_name = os.path.splitext(name)[0] + j
            if os.path.exists(img_path + img_name):
                break

        if i in train_list:
            ftrain.write(img_path + img_name + "\n")
        else:
            ftest.write(img_path + img_name + "\n")
    # 如果有只有圖片沒有xml的铭污，需要生成空白txt
    # imgs = os.listdir(img_path)
    # for img_name in imgs:
    #     txt_name = os.path.basename(img_name).split('.')[0] + '.txt'
    #     if not os.path.exists(label_path+txt_name):
    #         a = open(label_path+txt_name,'w')
    #     ftrain.write(img_path+img_name + "\n")

if __name__ == '__main__':
    main()

格式和 yolov5_v6.1 差不多

準(zhǔn)備如下，注意 labelTxt 名字

./yq_xz_left_right:
        ├── Annotations
        │   ├── 1.xml
        │   └── 2.xml
        ├── images
        │   ├── 1.jpg
        │   └── 2.jpg
        ├── labelTxt
        │   ├── 1.txt
        │   └── 2.txt
        ├── test.txt
        └── train.txt

開始訓(xùn)練

訓(xùn)練準(zhǔn)備

git clone https://github.com/hukaixuan19970627/yolov5_obb.git
cd yolov5_obb

修改 data/demo.yaml 里面的 train test 的 txt路徑

# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]

path: ./dataset   # dataset root dir
train: yq_xz_left_right/train.txt    # train images (relative to 'path') 
val: yq_xz_left_right/test.txt   # val images (relative to 'path') 
test: yq_xz_left_right/test.txt  # test images (optional)

# Classes
nc: 2  # number of classes
names: ['daozha','truck']  # class names

注意類別至少2類讥蟆，修改 data/hyp/xxx.yaml 學(xué)習(xí)率等參數(shù)

訓(xùn)練命令和測(cè)試命令

python -m torch.distributed.launch --nproc_per_node 3 train.py --device 1,2,3 --weights '' --cfg models_obb/yolov5x.yaml --data data/daozha2.yaml --hyp runs/train/daozhaX_lr00552/hyp.yaml --epochs 350 --batch-size 33 --imgsz 640 --name fuyuan --sync-bn

python detect.py --weights xxx --source xxx

一鍵順控模板準(zhǔn)備

# YOLOv5 ?? by Ultralytics, GPL-3.0 license
"""
Run inference on images, videos, directories, streams, etc.

Usage:
    $ python path/to/detect.py --weights yolov5s.pt --source 0  # webcam
                                                             img.jpg  # image
                                                             vid.mp4  # video
                                                             path/  # directory
                                                             path/*.jpg  # glob
                                                             'https://youtu.be/Zgi9g1ksQHc'  # YouTube
                                                             'rtsp://example.com/media.mp4'  # RTSP, RTMP, HTTP stream
"""

import argparse
import os
import sys
from pathlib import Path

import cv2
import torch
import torch.backends.cudnn as cudnn

FILE = Path(__file__).resolve()
ROOT = FILE.parents[0]  # YOLOv5 root directory
if str(ROOT) not in sys.path:
    sys.path.append(str(ROOT))  # add ROOT to PATH
ROOT = Path(os.path.relpath(ROOT, Path.cwd()))  # relative

from models.common import DetectMultiBackend
from utils.datasets import IMG_FORMATS, VID_FORMATS, LoadImages, LoadStreams
from utils.general import (LOGGER, check_file, check_img_size, check_imshow, check_requirements, colorstr,
                           increment_path, non_max_suppression, non_max_suppression_obb, print_args, scale_coords, scale_polys, strip_optimizer, xyxy2xywh)
from utils.plots import Annotator, colors, save_one_box
from utils.torch_utils import select_device, time_sync
from utils.rboxs_utils import poly2rbox, rbox2poly

def DeleteSmallBox(det):
    boxes = det.cpu().numpy()
    """
    in: 檢出超過3個(gè)框的det
    out: 最大的2個(gè)框
    """
    areas = list()
    RBoxes = list()
    for box in boxes:
        w,h = box[2],box[3]
        areas.append(w*h)
    print('AREAS >>',areas)
    for i in range(2):
        print('RBOXES >>',RBoxes)
        RBoxes.append(boxes[areas.index(max(areas))])
        areas.pop(areas.index(max(areas)))
    W1 = RBoxes[0][2]
    W2 = RBoxes[1][2]
    if W1>W2:
        RBoxes[0],RBoxes[1] = RBoxes[1],RBoxes[0]
    return RBoxes

@torch.no_grad()
def run(weights=ROOT / 'yolov5s.pt',  # model.pt path(s)
        source=ROOT / 'data/images',  # file/dir/URL/glob, 0 for webcam
        imgsz=(640, 640),  # inference size (height, width)
        conf_thres=0.25,  # confidence threshold
        iou_thres=0.15,  # NMS IOU threshold
        max_det=1000,  # maximum detections per image
        device='',  # cuda device, i.e. 0 or 0,1,2,3 or cpu
        view_img=False,  # show results
        save_txt=False,  # save results to *.txt
        save_conf=False,  # save confidences in --save-txt labels
        save_crop=False,  # save cropped prediction boxes
        nosave=False,  # do not save images/videos
        classes=None,  # filter by class: --class 0, or --class 0 2 3
        agnostic_nms=False,  # class-agnostic NMS
        augment=False,  # augmented inference
        visualize=False,  # visualize features
        update=False,  # update all models
        project=ROOT / 'runs/detect',  # save results to project/name
        name='exp',  # save results to project/name
        exist_ok=False,  # existing project/name ok, do not increment
        line_thickness=3,  # bounding box thickness (pixels)
        hide_labels=False,  # hide labels
        hide_conf=False,  # hide confidences
        half=False,  # use FP16 half-precision inference
        dnn=False,  # use OpenCV DNN for ONNX inference
        ):
    source = str(source)
    save_img = not nosave and not source.endswith('.txt')  # save inference images
    is_file = Path(source).suffix[1:] in (IMG_FORMATS + VID_FORMATS)
    is_url = source.lower().startswith(('rtsp://', 'rtmp://', 'http://', 'https://'))
    webcam = source.isnumeric() or source.endswith('.txt') or (is_url and not is_file)
    if is_url and is_file:
        source = check_file(source)  # download

    # Directories
    save_dir = increment_path(Path(project) / name, exist_ok=exist_ok)  # increment run
    (save_dir / 'labels' if save_txt else save_dir).mkdir(parents=True, exist_ok=True)  # make dir

    # Load model
    device = select_device(device)
    model = DetectMultiBackend(weights, device=device, dnn=dnn)
    stride, names, pt, jit, onnx, engine = model.stride, model.names, model.pt, model.jit, model.onnx, model.engine
    imgsz = check_img_size(imgsz, s=stride)  # check image size

    # Half
    half &= (pt or jit or engine) and device.type != 'cpu'  # half precision only supported by PyTorch on CUDA
    if pt or jit:
        model.model.half() if half else model.model.float()

    # Dataloader
    if webcam:
        view_img = check_imshow()
        cudnn.benchmark = True  # set True to speed up constant image size inference
 
        # source = '/ssd/yangyuqian/projects/yolov5_obb-master/dataset/mp4/h/0026_normal.mp4'
        dataset = LoadStreams(source, img_size=imgsz, stride=stride, auto=pt)
        bs = len(dataset)  # batch_size
    else:
        dataset = LoadImages(source, img_size=imgsz, stride=stride, auto=pt)
        bs = 1  # batch_size
    vid_path, vid_writer = [None] * bs, [None] * bs

    # Run inference
    model.warmup(imgsz=(1, 3, *imgsz), half=half)  # warmup
    dt, seen = [0.0, 0.0, 0.0], 0
    for path, im, im0s, vid_cap, s in dataset:
        txt_name = path.replace('.mp4','.txt')
        t1 = time_sync()
        im = torch.from_numpy(im).to(device)
        im = im.half() if half else im.float()  # uint8 to fp16/32
        im /= 255  # 0 - 255 to 0.0 - 1.0
        if len(im.shape) == 3:
            im = im[None]  # expand for batch dim
        t2 = time_sync()
        dt[0] += t2 - t1

        # Inference
        visualize = increment_path(save_dir / Path(path).stem, mkdir=True) if visualize else False
        pred = model(im, augment=augment, visualize=visualize)
        t3 = time_sync()
        dt[1] += t3 - t2

        # NMS
        # pred: list*(n, [cxcylsθ, conf, cls]) θ ∈ [-pi/2, pi/2)
        pred = non_max_suppression_obb(pred, conf_thres, iou_thres, classes, agnostic_nms, multi_label=True, max_det=max_det)
        dt[2] += time_sync() - t3

        # Second-stage classifier (optional)
        # pred = utils.general.apply_classifier(pred, classifier_model, im, im0s)

        # Process predictions
        need_check = False
        for i, det in enumerate(pred):  # per image
            # print(len(det),len(pred))
            if len(det)>2:
                res = DeleteSmallBox(det) # 取其中的兩個(gè)大框
                need_check = True
                print(path)
            elif len(det)<=1:
                print("number of isolator is less than 2!>>")
                continue
            else:
                res = det.cpu().numpy().tolist()
                W1 = res[0][0]
                W2 = res[1][0]
                if W1>W2:
                    print("調(diào)換位置")

                    res2 = [res[1],res[0]]
                    res = res2
            # print(det,'\n',res)
            with open(txt_name,'a+') as f:
                # if need_check:
                #     # print('檢查',path,W1,W2)
                # need_check = False
                f.write('%s,%s  >>>%s\n' %(res[0][4],res[1][4],str(i)))
            # continue
            pred_poly = rbox2poly(det[:, :5]) # (n, [x1 y1 x2 y2 x3 y3 x4 y4])
            seen += 1
            if webcam:  # batch_size >= 1
                p, im0, frame = path[i], im0s[i].copy(), dataset.count
                s += f'{i}: '
            else:
                p, im0, frame = path, im0s.copy(), getattr(dataset, 'frame', 0)

            p = Path(p)  # to Path
            save_path = str(save_dir / p.name)  # im.jpg
            txt_path = str(save_dir / 'labels' / p.stem) + ('' if dataset.mode == 'image' else f'_{frame}')  # im.txt
            s += '%gx%g ' % im.shape[2:]  # print string
            gn = torch.tensor(im0.shape)[[1, 0, 1, 0]]  # normalization gain whwh
            imc = im0.copy() if save_crop else im0  # for save_crop
            annotator = Annotator(im0, line_width=line_thickness, example=str(names))
            if len(det):
                # Rescale polys from img_size to im0 size
                # det[:, :4] = scale_coords(im.shape[2:], det[:, :4], im0.shape).round()
                pred_poly = scale_polys(im.shape[2:], pred_poly, im0.shape)
                det = torch.cat((pred_poly, det[:, -2:]), dim=1) # (n, [poly conf cls])

                # Print results
                for c in det[:, -1].unique():
                    n = (det[:, -1] == c).sum()  # detections per class
                    s += f"{n} {names[int(c)]}{'s' * (n > 1)}, "  # add to string

                # Write results
                for *poly, conf, cls in reversed(det):
                    if save_txt:  # Write to file
                        # xywh = (xyxy2xywh(torch.tensor(xyxy).view(1, 4)) / gn).view(-1).tolist()  # normalized xywh
                        poly = poly.tolist()
                        line = (cls, *poly, conf) if save_conf else (cls, *poly)  # label format
                        with open(txt_path + '.txt', 'a') as f:
                            f.write(('%g ' * len(line)).rstrip() % line + '\n')

                    if save_img or save_crop or view_img:  # Add poly to image
                        c = int(cls)  # integer class
                        label = None if hide_labels else (names[c] if hide_conf else f'{names[c]} {conf:.2f}')
                        # annotator.box_label(xyxy, label, color=colors(c, True))
                        annotator.poly_label(poly, label, color=colors(c, True))
                        if save_crop: # Yolov5-obb doesn't support it yet
                            # save_one_box(xyxy, imc, file=save_dir / 'crops' / names[c] / f'{p.stem}.jpg', BGR=True)
                            pass

            # Print time (inference-only)
            LOGGER.info(f'{s}Done. ({t3 - t2:.3f}s)')

            # Stream results
            im0 = annotator.result()
            if view_img:
                cv2.imshow(str(p), im0)
                cv2.waitKey(1)  # 1 millisecond

            # Save results (image with detections)
            if save_img:
                if dataset.mode == 'image':
                    cv2.imwrite(save_path, im0)
                else:  # 'video' or 'stream'
                    if vid_path[i] != save_path:  # new video
                        vid_path[i] = save_path
                        if isinstance(vid_writer[i], cv2.VideoWriter):
                            vid_writer[i].release()  # release previous video writer
                        if vid_cap:  # video
                            fps = vid_cap.get(cv2.CAP_PROP_FPS)
                            w = int(vid_cap.get(cv2.CAP_PROP_FRAME_WIDTH))
                            h = int(vid_cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
                        else:  # stream
                            fps, w, h = 30, im0.shape[1], im0.shape[0]
                            save_path += '.mp4'
                        vid_writer[i] = cv2.VideoWriter(save_path, cv2.VideoWriter_fourcc(*'mp4v'), fps, (w, h))
                    vid_writer[i].write(im0)

    # Print results
    t = tuple(x / seen * 1E3 for x in dt)  # speeds per image
    LOGGER.info(f'Speed: %.1fms pre-process, %.1fms inference, %.1fms NMS per image at shape {(1, 3, *imgsz)}' % t)
    if save_txt or save_img:
        s = f"\n{len(list(save_dir.glob('labels/*.txt')))} labels saved to {save_dir / 'labels'}" if save_txt else ''
        LOGGER.info(f"Results saved to {colorstr('bold', save_dir)}{s}")
    if update:
        strip_optimizer(weights)  # update model (to fix SourceChangeWarning)


def parse_opt():
    parser = argparse.ArgumentParser()
    parser.add_argument('--weights', nargs='+', type=str, default=ROOT / 'runs/train/yolov5m_finetune/weights/best.pt', help='model path(s)')
    parser.add_argument('--source', type=str, default='dataset/dataset_demo_rate1.0_split1024_gap200/images/', help='file/dir/URL/glob, 0 for webcam')
    parser.add_argument('--imgsz', '--img', '--img-size', nargs='+', type=int, default=[640,640], help='inference size h,w')
    parser.add_argument('--conf-thres', type=float, default=0.3, help='confidence threshold')
    parser.add_argument('--iou-thres', type=float, default=0.4, help='NMS IoU threshold')
    parser.add_argument('--max-det', type=int, default=1000, help='maximum detections per image')
    parser.add_argument('--device', default='1', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
    parser.add_argument('--view-img', action='store_true', help='show results')
    parser.add_argument('--save-txt', action='store_true', help='save results to *.txt')
    parser.add_argument('--save-conf', action='store_true', help='save confidences in --save-txt labels')
    parser.add_argument('--save-crop', action='store_true', help='save cropped prediction boxes')
    parser.add_argument('--nosave', action='store_true', help='do not save images/videos')
    parser.add_argument('--classes', nargs='+', type=int, help='filter by class: --classes 0, or --classes 0 2 3')
    parser.add_argument('--agnostic-nms', action='store_true', help='class-agnostic NMS')
    parser.add_argument('--augment', action='store_true', help='augmented inference')
    parser.add_argument('--visualize', action='store_true', help='visualize features')
    parser.add_argument('--update', action='store_true', help='update all models')
    parser.add_argument('--project', default=ROOT / 'runs/detect', help='save results to project/name')
    parser.add_argument('--name', default='exp', help='save results to project/name')
    parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
    parser.add_argument('--line-thickness', default=2, type=int, help='bounding box thickness (pixels)')
    parser.add_argument('--hide-labels', default=False, action='store_true', help='hide labels')
    parser.add_argument('--hide-conf', default=False, action='store_true', help='hide confidences')
    parser.add_argument('--half', action='store_true', help='use FP16 half-precision inference')
    parser.add_argument('--dnn', action='store_true', help='use OpenCV DNN for ONNX inference')
    opt = parser.parse_args()
    opt.imgsz *= 2 if len(opt.imgsz) == 1 else 1  # expand
    print_args(FILE.stem, opt)
    return opt


def main(opt):
    check_requirements(exclude=('tensorboard', 'thop'))
    run(**vars(opt))


if __name__ == "__main__":
    opt = parse_opt()
    main(opt)

python detect.py --weights xxx --source /path/

MP4對(duì)應(yīng)的文件夾下面會(huì)生成對(duì)應(yīng)的 xxx.txt

generate_config.py
"""
生成template_config.ini
"""


import os
import configparser
config = configparser.ConfigParser()
# with open('./template_config.ini','w') as f:
#     pass
path = './template_config_fy.ini'
if os.path.exists(path):
    os.remove(path)
config.read(path)

def getallfiles(path):
    allfile = []
    file_xml = []
    for dirpath, dirnames, filenames in os.walk(path):
        for dir in dirnames:
            allfile.append(os.path.join(dirpath, dir))
        for name in filenames:
            allfile.append(os.path.join(dirpath, name))
    for file in allfile:
        if file.endswith('.tx',-4,-1):
            file_xml.append(file)
    return file_xml

txt_paths = getallfiles('./dataset/mp4/fy')

for txt_path in txt_paths:
    # print(txt_path)
    with open(txt_path,'r') as f:
        content = f.readlines()
    mid = int(len(content)/2)
    tstart = content[0].replace("  >>>0",'')
    tmid = content[mid].replace("  >>>0",'')
    tend = content[-1].replace("  >>>0",'')
    bn = os.path.basename(txt_path).split('.')[0]
    config.add_section(bn)
    # print(config[bn])
    # print(tstart,tmid,tend)
    # print(type(bn))
    if "k" in txt_path:
        type = 'open'
    else:
        type = 'close'
    for cont in content :
        config.set(bn, 'tstart',tstart.replace('\n',''))
        config.set(bn, 'tmid',tmid.replace('\n',''))
        config.set(bn, 'tend',tend.replace('\n',''))
        config.set(bn, 'type',type.replace('\n',''))
        config.set(bn, 'threshold','0.09')
config.write(open(path,'w'))