使用annotation-tool线段标注工具生成txt格式的标注文件
1、将所有txt文件合并成一个json文件(代码来自实验室师姐)
import cv2
import sys
import numpy as np
import os
from glob import glob
from math import cos, sin
from os.path import isfile, splitext, basename, isdir
from os import makedirs
import json
img_files_path = r'E:\Desktop\zhijiang_total\images'
img_files = glob('%s/*.txt' % img_files_path)
writePath = './train.json'
with open(writePath,'w') as fo:
label = []
for i in range(len(img_files)):
dic = {}
dic["filename"] = img_files[i][len(img_files_path)+1:-4]+'.jpg'
print(img_files[i][:-4]+'.jpg')
im = cv2.imread(img_files[i][:-4]+'.jpg')
h,w,_ = im.shape
# print(w,h)
lines = []
with open(img_files[i],'r') as fi:
lines = []
for line in fi:
line = line.split(',')
# print(line)
# print(img_files[i][len(img_files_path)+1:])
# print(int(len(line)/4))
# for j in range(int(len(line)/4)):
# t = 4*j
lines.append([int(float(line[1])*w),int(float(line[3])*h),int(float(line[2])*w),int(float(line[4])*h)])
# print(lines)
# os.system('pause')
dic["lines"] = lines
dic["height"] = h
dic["width"] = w
label.append(dic)
json.dump(label, fo)
2、运行
python dataset/wireframe.py <src-path> <dest-path>将我们自己的标注数据和图片转换成模型可训练的数据格式
3、运行
python ./train.py -d 0 --identifier baseline config/wireframe.yaml开始训练模型
补充：
标注工具annotation-tool（代码来自实验室师哥）
'''
Annotation tool
Tool used to annotate labels for LP detection. Main keys:
C: creates a new shape
A: creates a new vertex over mouse position
D: delete last vertex
S: change position of closest vertex (to mouse pos.)
X: switch to closest shape
N or enter: next picture
P: previous picture
For more commands, take a look at the main part...
Usage:
python annotation-tool.py max_width max_height file1 ... fileN
Obs. This tool is not fully tested and crashes under unexpetected situations.
If you find any and correct, feel free to send a pull request =)
'''
import cv2
import sys
import numpy as np
import os
from glob import glob
from math import cos, sin
from os.path import isfile, splitext, basename, isdir
from os import makedirs
def getRectPts(tlx,tly,brx,bry):
return np.matrix([[tlx,brx,brx,tlx],[tly,tly,bry,bry],[1.,1.,1.,1.]],dtype=float)
def image_files_from_folder(folder,upper=True):
extensions = ['jpg','jpeg','png']
img_files = []
for ext in extensions:
img_files += glob('%s/*.%s' % (folder,ext))
if upper:
img_files += glob('%s/*.%s' % (folder,ext.upper()))
return img_files
class Label:
def __init__(self,cl=-1,tl=np.array([0.,0.]),br=np.array([0.,0.]),prob=None):
self.__tl = tl
self.__br = br
self.__cl = cl
self.__prob = prob
def __str__(self):
return 'Class: %d, top_left(x:%f,y:%f), bottom_right(x:%f,y:%f)' % (self.__cl, self.__tl[0], self.__tl[1], self.__br[0], self.__br[1])
def copy(self):
return Label(self.__cl,self.__tl,self.__br)
def wh(self): return self.__br-self.__tl
def cc(self): return self.__tl + self.wh()/2
def tl(self): return self.__tl
def br(self): return self.__br
def tr(self): return np.array([self.__br[0],self.__tl[1]])
def bl(self): return np.array([self.__tl[0],self.__br[1]])
def cl(self): return self.__cl
def area(self): return np.prod(self.wh())
def prob(self): return self.__prob
def set_class(self,cl):
self.__cl = cl
def set_tl(self,tl):
self.__tl = tl
def set_br(self,br):
self.__br = br
def set_wh(self,wh):
cc = self.cc()
self.__tl = cc - .5*wh
self.__br = cc + .5*wh
def set_prob(self,prob):
self.__prob = prob
class Shape():
def __init__(self,pts=np.zeros((2,0)),max_sides=2,text=''):
self.pts = pts
self.max_sides = max_sides
self.text = text
def isValid(self):
# print(self.pts.shape)
return self.pts.shape[1] > 1
def write(self,fp):
fp.write('%d,' % self.pts.shape[1])
ptsarray = self.pts.flatten()
fp.write(''.join([('%f,' % value) for value in ptsarray]))
fp.write('%s,' % self.text)
fp.write('\n')
def read(self,line):
data = line.strip().split(',')
ss = int(data[0])
values = data[1:(ss*2 + 1)]
text = data[(ss*2 + 1)] if len(data) >= (ss*2 + 2) else ''
self.pts = np.array([float(value) for value in values]).reshape((2,ss))
self.text = text
def readShapes(path,obj_type=Shape):
shapes = []
with open(path) as fp:
for line in fp:
shape = obj_type()
shape.read(line)
shapes.append(shape)
return shapes
def writeShapes(path,shapes):
if len(shapes):
# print
with open(path,'w') as fp:
for shape in shapes:
# print(shape.pts)
if shape.isValid():
shape.write(fp)
def find_T_matrix(pts,t_pts):
A = np.zeros((8,9))
for i in range(0,4):
xi = pts[:,i];
xil = t_pts[:,i];
xi = xi.T
A[i*2, 3:6] = -xil[2]*xi
A[i*2, 6: ] = xil[1]*xi
A[i*2+1, :3] = xil[2]*xi
A[i*2+1, 6: ] = -xil[0]*xi
[U,S,V] = np.linalg.svd(A)
H = V[-1,:].reshape((3,3))
return H
def perspective_transform(wh,angles=np.array([0.,0.,0.]),zcop=1000., dpp=1000.):
rads = np.deg2rad(angles)
a = rads[0]; Rx = np.matrix([[1, 0, 0] , [0, cos(a), sin(a)] , [0, -sin(a), cos(a)] ])
a = rads[1]; Ry = np.matrix([[cos(a), 0, -sin(a)] , [0, 1, 0] , [sin(a), 0, cos(a)] ])
a = rads[2]; Rz = np.matrix([[cos(a), sin(a), 0] , [-sin(a), cos(a), 0] , [0, 0, 1] ])
R = Rx*Ry*Rz;
(w,h) = tuple(wh)
xyz = np.matrix([[0,0,w,w],[0,h,0,h],[0,0,0,0]])
hxy = np.matrix([[0,0,w,w],[0,h,0,h],[1,1,1,1]])
xyz = xyz - np.matrix([[w],[h],[0]])/2.
xyz = R*xyz
xyz = xyz - np.matrix([[0],[0],[zcop]])
hxyz = np.concatenate([xyz,np.ones((1,4))])
P = np.matrix([[1,0,0,0],[0,1,0,0],[0,0,-1./dpp,0]])
_hxy = P*hxyz
_hxy = _hxy/_hxy[2,:]
_hxy = _hxy + np.matrix([[w],[h],[0]])/2.
return find_T_matrix(hxy,_hxy)
def getWH(shape):
return np.array(shape[1::-1]).astype(float)
class ShapeDisplay(Shape):
def appendSide(self,pt):
if (self.max_sides == 0) or (self.pts.shape[1] < self.max_sides):
self.pts = np.append(self.pts,pt,axis=1)
def removeLast(self):
self.pts = self.pts[...,:-1]
def changeClosest(self,pt):
try:
idx = np.argmin(self.distanceTo(pt))
self.pts[...,idx] = pt[...,0]
except ValueError as e:
print("except : ", e)
def distanceTo(self,pt):
return np.sqrt(np.power((self.pts - pt),2).sum(0))
def shiftPts(self):
if self.pts.shape[1] > 1:
idx = range(1,self.pts.shape[1]) + [0]
self.pts = self.pts[...,idx]
def getSquare(self):
tl,br = self.pts.min(1),self.pts.max(1)
return Label(-1,tl,br)
def draw(self,drawLineFunc,drawCircleFunc,drawTextFunc,color=(255,255,255),txtcolor=(255,255,255)):
ss = self.pts.shape[1]
if ss:
text = self.text if len(self.text) else '.'
drawTextFunc(text,self.pts.min(1),color=txtcolor)
if ss > 1:
for i in range(ss):
drawLineFunc(self.pts[:,i],self.pts[:,(i+1)%ss],color=color)
drawCircleFunc(self.pts[:,0] ,color=(255-color[0],0,255-color[2]))
drawCircleFunc(self.pts[:,-1],color=(255-color[0],255-color[1],255-color[2]))
def rotation_transform(wh,angles=np.array([0.,0.,0.]),zcop=1000., dpp=1000.):
rads = np.deg2rad(angles)
a = rads[0]; Rx = np.matrix([[1, 0, 0] , [0, cos(a), sin(a)] , [0, -sin(a), cos(a)] ])
a = rads[1]; Ry = np.matrix([[cos(a), 0, -sin(a)] , [0, 1, 0] , [sin(a), 0, cos(a)] ])
a = rads[2]; Rz = np.matrix([[cos(a), sin(a), 0] , [-sin(a), cos(a), 0] , [0, 0, 1] ])
R = Rx*Ry*Rz
return R
class Display():
def __init__(self,I,width,height,wname='Display'):
try:
self.Iorig = I.copy()
self.width = width
self.height = height
self.wname = wname
self.Idisplay = self.Iorig.copy()
self.IdisplayCopy = self.Idisplay.copy()
self.reset_view()
self._setPerspective()
cv2.namedWindow(self.wname)
cv2.moveWindow(self.wname, 0, 0)
cv2.setMouseCallback(self.wname,self.mouse_callback)
except AttributeError as e:
print("except : ", e)
def reset_view(self):
self.cx, self.cy = .5,.5
wh = np.array([self.width,self.height],dtype=float)
self.zoom_factor = (wh/getWH(self.Iorig.shape)).min()
self.mouse_center = np.array([.5,.5])
self.angles = np.array([0.,0.,0.])
self._setPerspective()
def updatePerspectiveMatrix(self):
zf = self.zoom_factor
w,h = getWH(self.Iorig.shape)
self.dx = self.cx*w*zf - self.width/2.
self.dy = self.cy*h*zf - self.height/2.
R = np.eye(3)
R = np.matmul(R,np.matrix([[zf,0,-self.dx],[0,zf,-self.dy],[0,0,1]],dtype=float))
R = np.matmul(R,perspective_transform((w,h),angles=self.angles))
self.R = R
self.Rinv = np.linalg.inv(R)
def show(self):
cv2.imshow(self.wname,self.Idisplay)
def setPerspectiveAngle(self,addx=0.,addy=0.,addz=0.):
self.angles += np.array([addx,addy,addz])
self._setPerspective()
def _setPerspective(self,update=True):
if update:
self.updatePerspectiveMatrix()
self.IdisplayCopy = cv2.warpPerspective(self.Iorig,self.R,(self.width,self.height),borderValue=.0,flags=cv2.INTER_LINEAR)
def resetDisplay(self):
self.Idisplay = self.IdisplayCopy.copy()
def getMouseCenterRelative(self):
return self.mouse_center.copy().reshape((2,1))
def waitKey(self,time=50):
return cv2.waitKey(50) & 0x0000000FF
def __pt2xy(self,pt):
pt = np.squeeze(np.array(np.matmul(self.R,np.append(pt*getWH(self.Iorig.shape),1.))))
pt = pt[:2]/pt[2]
return tuple(pt.astype(int).tolist())
def __pts2xys(self,pts):
N = pts.shape[1]
pts = pts*getWH(self.Iorig.shape).reshape((2,1))
pts = np.concatenate((pts,np.ones((1,N))))
pts = np.squeeze(np.array(np.matmul(self.R,pts)))
pts = pts[:2]/pts[2]
return pts
def drawLine(self,pt1,pt2,color=(255,255,255),thickness=3):
pt1 = self.__pt2xy(pt1)
pt2 = self.__pt2xy(pt2)
cv2.line(self.Idisplay,pt1,pt2,color=color,thickness=thickness)
def drawCircle(self,center,color=(255,255,0),radius=7):
center = self.__pt2xy(center)
cv2.circle(self.Idisplay,center,radius,color,thickness=-1)
def drawText(self,text,bottom_left_pt,color=(255,255,0),bgcolor=(0,0,0),font_size=1):
bl_corner = self.__pt2xy(bottom_left_pt)
font = cv2.FONT_HERSHEY_SIMPLEX
wh_text,v = cv2.getTextSize(text, font, font_size, 3)
tl_corner = (bl_corner[0],bl_corner[1]-wh_text[1])
br_corner = (bl_corner[0]+wh_text[0],bl_corner[1])
cv2.rectangle(self.Idisplay, tl_corner, br_corner, bgcolor,-1)
cv2.putText(self.Idisplay,text,bl_corner,font,font_size,color,3)
def zoom(self,ff):
self.zoom_factor *= ff
self.cx,self.cy = self.mouse_center.tolist()
self._setPerspective()
def rectifyToPts(self,pts):
if pts.shape[1] != 4:
return
ptsh = pts*getWH(self.Iorig.shape).reshape((2,1))
ptsh = np.concatenate((ptsh,np.ones((1,4))))
to_pts = self.__pts2xys(pts)
wi,hi = (to_pts.min(1)[:2]).tolist()
wf,hf = (to_pts.max(1)[:2]).tolist()
to_pts = np.matrix([[wi,wf,wf,wi],[hi,hi,hf,hf],[1,1,1,1]])
self.R = find_T_matrix(ptsh,to_pts)
self.Rinv = np.linalg.inv(self.R)
self._setPerspective(update=False)
def mouse_callback(self,event,x,y,flags,param):
mc = np.array([x,y],dtype=float)
mc = np.matmul(self.Rinv,np.append(mc,1.))
mc = np.squeeze(np.array(mc))
self.mouse_center = (mc[:2]/mc[2])/getWH(self.Iorig.shape)
def selectClosest(shapes,pt):
if len(shapes):
mindist,selected = shapes[0].distanceTo(pt).min(),0
try:
for i,shape in enumerate(shapes[1:]):
shpdist = shape.distanceTo(pt).min()
if mindist > shpdist:
selected = i+1
mindist = shpdist
except ValueError as e:
print("except : ", e)
return selected
else:
return -1
def displayAllShapes(disp,shapes,selected,typing_mode):
for i,shape in enumerate(shapes):
color = (0,255,255) if i != selected else (0,255,0)
txtcolor = (0,0,255) if (i == selected and typing_mode) else (0,255,255)
shape.draw(disp.drawLine,disp.drawCircle,disp.drawText,color=color,txtcolor=txtcolor)
if __name__ == '__main__':
# if len(sys.argv) < 4:
# print __doc__
# sys.exit()
maxW = 1500 #int(sys.argv[1])
maxH = 900 #int(sys.argv[2])
# img_files = sys.argv[3:]
img_files_path = './new/'#############记得改文件名
img_files = glob('%s/*.*g' % img_files_path)
maxwh = np.array([maxW,maxH],dtype=float)
wname = 'Display'
# Key ids
ENTER = 10
ESC = 27
BACKSPACE = 8
ARROW_UP = 82
ARROW_DOWN = 84
ARROW_LEFT = 81
ARROW_RIGHT = 83
GREATER_THAN = 46
LESS_THAN = 44
HOME = 80
key_exit = ESC
key_next = [ord('n'),ENTER]
key_previous = ord('p')
key_zoom_in = ord('q')
key_zoom_out = ord('w')
key_append_vertex = ord('a')
key_remove_last_vertex = ord('d')
key_change_closest_vertex = ord('s')
key_create_new_shape = ord('c')
key_select_closest_shape = ord('x')
key_shift_pts = ord('g')
key_typing_mode = ord(' ')
key_delete_selected_shape = [ord('r')]
key_pitch_increase = ARROW_DOWN
key_pitch_decrease = ARROW_UP
key_yaw_increase = ARROW_LEFT
key_yaw_decrease = ARROW_RIGHT
key_roll_increase = GREATER_THAN
key_roll_decrease = LESS_THAN
key_perspective_reset = HOME
zoom_factor = 1.
action_keys = [key_exit,key_previous] + key_next
curr_image = 0
# print(len(img_files))
while curr_image < len(img_files):
img_file = img_files[curr_image]
print(f"current picture : {img_file}")
lab_file = splitext(img_file)[0] + '.txt'
if isfile(lab_file):
shapes = readShapes(lab_file,obj_type=ShapeDisplay)
selected = len(shapes) - 1
# print('t',shapes,shapes[selected],selected)
else:
shapes,selected = [ShapeDisplay()],0
# print('f',shapes,shapes[selected],selected)
# print(len(shapes))
zoom_factor = 1.
disp = Display(cv2.imread(img_file),maxW,maxH)
disp.show()
key = disp.waitKey()
typing_mode = False
# print('\taction_keys',action_keys)
# while True:
while not key in action_keys:
disp.resetDisplay()
displayAllShapes(disp,shapes,selected,typing_mode)
disp.show()
key = disp.waitKey(10)
if typing_mode:
if key == key_typing_mode:
typing_mode = False
else:
if key != 255:
if key >= 176:
key = key - 176 + 48
if key == BACKSPACE: # backspace
shapes[selected].text = shapes[selected].text[:-1]
else:
shapes[selected].text += str(chr(key)).upper()
key = 255
continue
if key == key_zoom_in:
disp.zoom(1.5)
if key == key_zoom_out:
disp.zoom(.5)
if key == key_yaw_increase:
disp.setPerspectiveAngle(addy=10.)
if key == key_yaw_decrease:
disp.setPerspectiveAngle(addy=-10.)
if key == key_pitch_increase:
disp.setPerspectiveAngle(addx=10.)
if key == key_pitch_decrease:
disp.setPerspectiveAngle(addx=-10.)
if key == key_roll_increase:
disp.setPerspectiveAngle(addz=10.)
if key == key_roll_decrease:
disp.setPerspectiveAngle(addz=-10.)
if key == key_perspective_reset:
disp.reset_view()
# print(shapes,len(shapes))
if len(shapes):
# print(shapes[selected].pts)
if key == ord('l'):
disp.rectifyToPts(shapes[selected].pts)
if key == key_typing_mode:
typing_mode = True
if key == key_append_vertex:
print ('Append vertex')
shapes[selected].appendSide(disp.getMouseCenterRelative())
if key == key_remove_last_vertex:
print ('Remove last vertex')
shapes[selected].removeLast()
if key == key_change_closest_vertex:
print ('Change closest vertex')
shapes[selected].changeClosest(disp.getMouseCenterRelative())
if key in key_delete_selected_shape:
print ('Delete closest vertex')
del shapes[selected]
pt = disp.getMouseCenterRelative()
selected = selectClosest(shapes,pt)
if key == key_shift_pts:
shapes[selected].shiftPts()
if key == key_create_new_shape:
print ('Create new shape')
shapes.append(ShapeDisplay())
selected = len(shapes)-1
if key == key_select_closest_shape:
print ('Select closest')
pt = disp.getMouseCenterRelative()
selected = selectClosest(shapes,pt)
if key == key_exit:
sys.exit()
if key in ([key_previous] + key_next):
# print(shapes)
writeShapes(lab_file,shapes)
curr_image += 1 if key in key_next else -1
curr_image = max(curr_image,0)
continue
后续根据识别结果，进行二次标注处理
# ################################ 二次标注处理函数 #################################
# path 图片所在文件夹路径
# imName是图片名称
# boxes是线段坐标, 类型是列表, 元素为元组, 形式为[(y1,x1),(y2,x2)]
def secondaryLabeling(path, imgName, boxes):
imgPath = os.path.join(path, imgName)
txtPath = os.path.join(path, imgName.split('.')[0]+'.txt')
im = cv2.imread(imgPath)
h, w, d = im.shape
# print(f"w is {w}, h is {h}, d is {d}")
with open(txtPath, 'w') as fw:
for box in boxes:
fw.write('2,')
a, b = box
ptsarray = [round(a[1]/w, 6), round(b[1]/w, 6), round(a[0]/h, 6), round(b[0]/h, 6)]
fw.write(''.join([('%f,' % value) for value in ptsarray]))
fw.write(',\n')
扫一扫
377
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
