def dota2darknet(imgpath, txtpath, dstpath, extractclassname):
"""
:param imgpath: the path of images
:param txtpath: the path of txt in dota format
:param dstpath: the path of txt in YOLO format
:param extractclassname: the category you selected
:return:
"""
filelist = util.GetFileFromThisRootDir(txtpath)
for fullname in filelist:
objects = util.parse_dota_poly(fullname)
name = os.path.splitext(os.path.basename(fullname))[0]
img_fullname = os.path.join(imgpath, name + '.png')
img = Image.open(img_fullname)
img_w, img_h = img.size
# print img_w,img_h
with open(os.path.join(dstpath, name + '.txt'), 'w') as f_out:
for obj in objects:
poly = obj['poly']
bbox = np.array(util.dots4ToRecC(poly, img_w, img_h))
if (sum(bbox <= 0) + sum(bbox >= 1)) >= 1:
continue
if (obj['name'] in extractclassname):
id = extractclassname.index(obj['name'])
else:
continue
outline = str(id) + ' ' + ' '.join(list(map(str, bbox)))
f_out.write(outline + '\n')
def imageformatTrans(srcpath, dstpath, format):
filelist = util.GetFileFromThisRootDir(srcpath)
for fullname in filelist:
img = cv2.imread(fullname)
basename = util.custombasename(fullname)
dstname = os.path.join(dstpath, basename + format)
cv2.imwrite(dstname, img)
def DOTA2COCO(srcpath, destfile):
imageparent = os.path.join(srcpath, 'images')
labelparent = os.path.join(srcpath, 'labelTxt')
data_dict = {}
info = {
'contributor': 'captain group',
'data_created': '2019',
'description': 'This is 1.5 version of DOTA dataset.',
'url': 'http://captain.whu.edu.cn/DOTAweb/',
'version': '1.5',
'year': 2019
}
data_dict['info'] = info
data_dict['images'] = []
data_dict['categories'] = []
data_dict['annotations'] = []
for idex, name in enumerate(wordname_16):
single_cat = {'id': idex + 1, 'name': name, 'supercategory': name}
data_dict['categories'].append(single_cat)
inst_count = 1
image_id = 1
with open(destfile, 'w') as f_out:
filenames = util.GetFileFromThisRootDir(labelparent)
for file in filenames:
basename = util.custombasename(file)
# image_id = int(basename[1:])
imagepath = os.path.join(imageparent, basename + '.png')
img = cv2.imread(imagepath)
height, width, c = img.shape
single_image = {}
single_image['file_name'] = basename + '.png'
single_image['id'] = image_id
single_image['width'] = width
single_image['height'] = height
data_dict['images'].append(single_image)
# annotations
objects = util.parse_dota_poly2(file)
for obj in objects:
single_obj = {}
single_obj['area'] = obj['area']
single_obj['category_id'] = wordname_16.index(obj['name']) + 1
single_obj['segmentation'] = []
single_obj['segmentation'].append(obj['poly'])
single_obj['iscrowd'] = 0
xmin, ymin, xmax, ymax = min(obj['poly'][0::2]), min(obj['poly'][1::2]), \
max(obj['poly'][0::2]), max(obj['poly'][1::2])
width, height = xmax - xmin, ymax - ymin
single_obj['bbox'] = xmin, ymin, width, height
single_obj['image_id'] = image_id
data_dict['annotations'].append(single_obj)
single_obj['id'] = inst_count
inst_count = inst_count + 1
image_id = image_id + 1
json.dump(data_dict, f_out)
def mergebase_parallel(srcpath, dstpath, nms, nms_thresh):
pool = Pool(16)
filelist = util.GetFileFromThisRootDir(srcpath)
mergesingle_fn = partial(mergesingle, dstpath, nms, nms_thresh)
# pdb.set_trace()
pool.map(mergesingle_fn, filelist)
def DOTA2COCOTest(srcpath, destfile, cls_names, ext='.png'):
imageparent = os.path.join(srcpath, 'images')
data_dict = {}
data_dict['images'] = []
data_dict['categories'] = []
for idex, name in enumerate(cls_names):
single_cat = {'id': idex + 1, 'name': name, 'supercategory': name}
data_dict['categories'].append(single_cat)
image_id = 1
with open(destfile, 'w') as f_out:
filenames = util.GetFileFromThisRootDir(imageparent)
for file in filenames:
basename = util.custombasename(file)
imagepath = os.path.join(imageparent, basename + ext)
img = Image.open(imagepath)
height = img.height
width = img.width
single_image = {}
single_image['file_name'] = basename + ext
single_image['id'] = image_id
single_image['width'] = width
single_image['height'] = height
data_dict['images'].append(single_image)
image_id = image_id + 1
json.dump(data_dict, f_out)
def DOTA2COCOTrain(srcpath, destfile, cls_names, difficult='2'):
# set difficult to filter '2', '1', or do not filter, set '-1'
imageparent = os.path.join(srcpath, 'images')
labelparent = os.path.join(srcpath, 'labelTxt-v1.0')
data_dict = {}
data_dict['images'] = []
data_dict['categories'] = []
data_dict['annotations'] = []
for idex, name in enumerate(cls_names):
single_cat = {'id': idex + 1, 'name': name, 'supercategory': name}
data_dict['categories'].append(single_cat)
inst_count = 1
image_id = 1
with open(destfile, 'w') as f_out:
filenames = util.GetFileFromThisRootDir(labelparent)
for file in filenames:
basename = util.custombasename(file)
# image_id = int(basename[1:])
imagepath = os.path.join(imageparent, basename + '.png')
img = Image.open(imagepath)
height = img.height
width = img.width
# img = cv2.imread(imagepath)
# height, width, c = img.shape
single_image = {}
single_image['file_name'] = basename + '.png'
single_image['id'] = image_id
single_image['width'] = width
single_image['height'] = height
data_dict['images'].append(single_image)
# annotations
objects = util.parse_dota_poly2(file)
for obj in objects:
if obj['difficult'] == difficult:
print('difficult: ', difficult)
continue
single_obj = {}
single_obj['area'] = obj['area']
single_obj['category_id'] = cls_names.index(obj['name']) + 1
single_obj['segmentation'] = []
single_obj['segmentation'].append(obj['poly'])
single_obj['iscrowd'] = 0
xmin, ymin, xmax, ymax = min(obj['poly'][0::2]), min(obj['poly'][1::2]), \
max(obj['poly'][0::2]), max(obj['poly'][1::2])
width, height = xmax - xmin, ymax - ymin
single_obj['bbox'] = xmin, ymin, width, height
single_obj['image_id'] = image_id
data_dict['annotations'].append(single_obj)
single_obj['id'] = inst_count
inst_count = inst_count + 1
image_id = image_id + 1
json.dump(data_dict, f_out)
def splitdata(self, rate):
imagelist = util.GetFileFromThisRootDir(self.srcpath)
imagenames = [
util.custombasename(x) for x in imagelist
if (util.custombasename(x) != 'Thumbs')
]
for name in imagenames:
self.SplitSingle(name, rate, self.ext)
def splitdata(self, rate):
imagelist = util.GetFileFromThisRootDir(self.srcpath)
imagenames = [util.custombasename(x) for x in imagelist if (util.custombasename(x) != 'Thumbs')]
# worker = partial(self.SplitSingle, rate=rate, extent=self.ext)
worker = partial(split_single_warp, split_base=self, rate=rate, extent=self.ext)
self.pool.map(worker, imagenames)
def rotate(srcpath, dstpath, num_process=16):
pool = Pool(num_process)
imgnames = util.GetFileFromThisRootDir(os.path.join(srcpath, 'images'))
names = [util.custombasename(x) for x in imgnames]
rotate_fun = partial(rotate_single_run, srcpath=srcpath, dstpath=dstpath)
pool.map(rotate_fun, names)
def delete(imgpath, txtpath):
filelist = util.GetFileFromThisRootDir(
txtpath) # fileist=['/.../P0005.txt', ..., /.../P000?.txt]
for fullname in filelist: # fullname='/.../P000?.txt'
name = os.path.splitext(os.path.basename(fullname))[0] # name='P000?'
img_fullname = os.path.join(imgpath, name +
'.png') # img_fullname='/.../P000?.png'
if not os.path.exists(img_fullname): # 如果文件bu存在
os.remove(fullname)
def __init__(self, basepath):
self.basepath = basepath
self.labelpath = os.path.join(basepath, 'labelTxt')
self.imagepath = os.path.join(basepath, 'images')
self.imgpaths = util.GetFileFromThisRootDir(self.labelpath)
self.imglist = [util.custombasename(x) for x in self.imgpaths]
self.catToImgs = defaultdict(list)
self.ImgToAnns = defaultdict(list)
self.createIndex()
def __init__(self, basepath):
self.basepath = basepath
self.labelpath = os.path.join(basepath, 'labelTxt')
self.imagepath = os.path.join(basepath, 'images')
self.imgpaths = util.GetFileFromThisRootDir(self.labelpath)
self.imglist = [util.custombasename(x) for x in self.imgpaths] # 去掉后缀
self.catToImgs = defaultdict(
list) # defaultdict 接受工厂方法，list 为[] int 为0 str 为"",set为 set()
self.ImgToAnns = defaultdict(list)
self.createIndex() # 创建种类到图片id 以及图像id 到对象列表的索引
def filecopy_v2(srcpath, dstpath, num_process=32):
filenames = util.GetFileFromThisRootDir(srcpath)
filenames = [os.path.basename(x.strip()) for x in filenames]
path_pair_list = []
for name in filenames:
srcdir = os.path.join(srcpath, name)
dstdir = os.path.join(dstpath, name)
path_pair_list.append((srcdir, dstdir))
copy_pool = Pool(num_process)
copy_pool.map(filecopy_single, path_pair_list)
def drawLongsideFormatimg(imgpath,
txtpath,
dstpath,
extractclassname,
thickness=2):
"""
根据labels绘制边框(label_format:classid, x_c_normalized, y_c_normalized, longside_normalized, shortside_normalized, Θ)
:param imgpath: the path of images
:param txtpath: the path of txt in longside format
:param dstpath: the path of image_drawed
:param extractclassname: the category you selected
"""
if os.path.exists(dstpath):
shutil.rmtree(dstpath) # delete output folder
os.makedirs(dstpath) # make new output folder
# 设置画框的颜色 colors = [[178, 63, 143], [25, 184, 176], [238, 152, 129],....,[235, 137, 120]]随机设置RGB颜色
colors = [[random.randint(0, 255) for _ in range(3)]
for _ in range(len(extractclassname))]
filelist = util.GetFileFromThisRootDir(
txtpath) # fileist=['/.../P0005.txt', ..., /.../P000?.txt]
for fullname in filelist: # fullname='/.../P000?.txt'
objects = util.parse_longsideformat(fullname)
'''
objects[i] = [classid, x_c_normalized, y_c_normalized, longside_normalized, shortside_normalized, theta]
'''
name = os.path.splitext(os.path.basename(fullname))[0] # name='P000?'
img_fullname = os.path.join(imgpath, name +
'.png') # img_fullname='/.../P000?.png'
img_savename = os.path.join(dstpath, name +
'_.png') # img_fullname='/.../_P000?.png'
img = Image.open(img_fullname) # 图像被打开但未被读取
img_w, img_h = img.size
img = cv2.imread(img_fullname) # 读取图像像素
for i, obj in enumerate(objects):
# obj = [classid, x_c_normalized, y_c_normalized, longside_normalized, shortside_normalized, float:0-179]
class_index = obj[0]
# rect=[(x_c,y_c),(w,h),Θ] Θ:flaot[0-179] -> (-180,0)
rect = longsideformat2cvminAreaRect(obj[1], obj[2], obj[3], obj[4],
(obj[5] - 179.9))
# poly = [(x1,y1),(x2,y2),(x3,y3),(x4,y4)]
poly = np.float32(cv2.boxPoints(rect)) # 返回rect对应的四个点的值 normalized
# 四点坐标反归一化 取整
poly[:, 0] = poly[:, 0] * img_w
poly[:, 1] = poly[:, 1] * img_h
poly = np.int0(poly)
# 画出来
cv2.drawContours(image=img,
contours=[poly],
contourIdx=-1,
color=colors[int(class_index)],
thickness=thickness)
cv2.imwrite(img_savename, img)
def Aerial2COCOTrain(srcpath, destfile, cls_names):
imageparent = os.path.join(srcpath, 'images')
labelparent = os.path.join(srcpath, 'labelTxt')
data_dict = {}
data_dict['images'] = []
data_dict['categories'] = []
data_dict['annotations'] = []
for idex, name in enumerate(cls_names):
single_cat = {'id': idex + 1, 'name': name, 'supercategory': name}
data_dict['categories'].append(single_cat)
inst_count = 1
image_id = 1
with open(destfile, 'w') as f_out:
filenames = util.GetFileFromThisRootDir(labelparent)
for file in filenames:
basename = util.custombasename(file)
imagepath = os.path.join(imageparent, basename + '.tif')
img = cv2.imread(imagepath)
height, width, c = img.shape
single_image = {}
single_image['file_name'] = basename + '.tif'
single_image['id'] = image_id
single_image['width'] = width
single_image['height'] = height
data_dict['images'].append(single_image)
# annotations
objects = parse_aerial_poly2(file)
for obj in objects:
single_obj = {}
single_obj['area'] = obj['area']
single_obj['category_id'] = cls_names.index(obj['name']) + 1
single_obj['segmentation'] = []
single_obj['segmentation'].append(obj['poly'])
single_obj['iscrowd'] = 0
xmin, ymin, xmax, ymax = min(obj['poly'][0::2]), min(obj['poly'][1::2]), \
max(obj['poly'][0::2]), max(obj['poly'][1::2])
width, height = xmax - xmin, ymax - ymin
single_obj['bbox'] = xmin, ymin, width, height
single_obj['image_id'] = image_id
data_dict['annotations'].append(single_obj)
single_obj['id'] = inst_count
inst_count = inst_count + 1
image_id = image_id + 1
json.dump(data_dict, f_out, indent=2)
def mergebase(srcpath, dstpath, nms):
if not os.path.exists(dstpath):
os.makedirs(dstpath)
img_result = {}
filelist = util.GetFileFromThisRootDir(srcpath)
for fullname in filelist:
name = util.custombasename(fullname)
#print('name:', name)
dstname = os.path.join(dstpath, name + '.txt')
with open(fullname, 'r') as f_in:
nameboxdict = {}
lines = f_in.readlines()
splitlines = [x.strip().split(' ') for x in lines]
for splitline in splitlines:
subname = splitline[0]
splitname = subname.split('__')
oriname = splitname[0]
pattern1 = re.compile(r'__\d+___\d+')
#print('subname:', subname)
x_y = re.findall(pattern1, subname)
x_y_2 = re.findall(r'\d+', x_y[0])
x, y = int(x_y_2[0]), int(x_y_2[1])
pattern2 = re.compile(r'__([\d+\.]+)__\d+___')
rate = re.findall(pattern2, subname)[0]
confidence = splitline[1]
poly = list(map(float, splitline[2:]))
origpoly = poly2origpoly(poly, x, y, rate)
det = origpoly
det.append(confidence)
det = list(map(float, det))
if (oriname not in nameboxdict):
nameboxdict[oriname] = []
nameboxdict[oriname].append(det)
nameboxnmsdict = nmsbynamedict(nameboxdict, nms, nms_thresh)
with open(dstname, 'w') as f_out:
for imgname in nameboxnmsdict:
if (imgname not in img_result):
img_result[imgname] = {}
if (name not in img_result[imgname]):
img_result[imgname][name] = []
for det in nameboxnmsdict[imgname]:
img_result[imgname][name].append(det)
#print('det:', det)
confidence = det[-1]
bbox = det[0:-1]
outline = imgname + ' ' + str(
confidence) + ' ' + ' '.join(map(str, bbox))
#print('outline:', outline)
f_out.write(outline + '\n')
return img_result
def mergebase_parallel(srcpath, dstpath, nms):
"""
将源路径中所有的txt目标信息,经nms后存入目标路径中的同名txt
@param srcpath: 合并前信息保存的txt源路径
@param dstpath: 合并后信息保存的txt目标路径
@param nms: NMS函数
"""
pool = Pool(16)
filelist = util.GetFileFromThisRootDir(srcpath)
mergesingle_fn = partial(mergesingle, dstpath, nms)
# pdb.set_trace()
pool.map(mergesingle_fn, filelist)
def mergebase(srcpath, dstpath, nms):
"""
读入子图的检测结果，返回最终的检测结果,还是子图的检测结果名称还有下划线
:param srcpath:
:param dstpath:
:param nms:
:return:
"""
filelist = util.GetFileFromThisRootDir(srcpath)
for fullname in filelist:
name = util.custombasename(fullname) # 去掉拓展名
# print('name:', name)
dstname = os.path.join(dstpath, name + '.txt') #形成dstname 的目标文件
with open(fullname, 'r') as f_in:
nameboxdict = {}
lines = f_in.readlines()
splitlines = [x.strip().split(' ') for x in lines]
for splitline in splitlines:
subname = splitline[0] #子图的名称
splitname = subname.split('__')
oriname = splitname[0] # 原图的名称
pattern1 = re.compile(r'__\d+___\d+')
# print('subname:', subname)
x_y = re.findall(pattern1, subname) # 返回的是一个只包含一个元素的list，所以下一步[0]
x_y_2 = re.findall(r'\d+', x_y[0])
x, y = int(x_y_2[0]), int(x_y_2[1]) # 解出 x,y的值
pattern2 = re.compile(r'__([\d+\.]+)__\d+___')
rate = re.findall(pattern2, subname)[0] # 缩放比例
confidence = splitline[1] # 置信度
poly = list(map(float, splitline[2:])) # 解出多边形的点
origpoly = poly2origpoly(poly, x, y, rate) # 还原出原来的poly
det = origpoly
det.append(confidence) # [8个坐标，置信度]
det = list(map(float, det))
if (oriname not in nameboxdict):
nameboxdict[oriname] = []
nameboxdict[oriname].append(det) # 原图对应的多个检测结果
nameboxnmsdict = nmsbynamedict(nameboxdict, nms, nms_thresh) # 进行nms
with open(dstname, 'w') as f_out:
for imgname in nameboxnmsdict:
for det in nameboxnmsdict[imgname]:
# print('det:', det)
confidence = det[-1]
bbox = det[0:-1]
outline = imgname + ' ' + str(confidence) + ' ' + ' '.join(map(str, bbox))
# print('outline:', outline)
f_out.write(outline + '\n')
def mergebase_parallel(srcpath, dstpath, nms):
"""
多线程转换为原图
:param srcpath:
:param dstpath:
:param nms:
:return:
"""
pool = Pool(16)
filelist = util.GetFileFromThisRootDir(srcpath)
mergesingle_fn = partial(mergesingle, dstpath, nms)
# pdb.set_trace()
pool.map(mergesingle_fn, filelist)
def dota2Darknet(imgpath, txtpath, dstpath, extractclassname):
"""
:param imgpath: the path of images
:param txtpath: the path of txt in dota format
:param dstpath: the path of txt in YOLO format
:param extractclassname: the category you selected
:return:
txt format: id x y w h
"""
if os.path.exists(dstpath):
shutil.rmtree(dstpath) # delete output folder
os.makedirs(dstpath) # make new output folder
filelist = util.GetFileFromThisRootDir(
txtpath) # fileist=['/.../P0005.txt', ..., /.../P000?.txt]
for fullname in filelist: # fullname='/.../P000?.txt'
objects = util.parse_dota_poly(fullname)
'''
objects =
[{'name': 'ship',
'difficult': '1',
'poly': [(1054.0, 1028.0), (1063.0, 1011.0), (1111.0, 1040.0), (1112.0, 1062.0)],
'area': 1159.5
},
...
]
'''
name = os.path.splitext(os.path.basename(fullname))[0] # name='P000?'
img_fullname = os.path.join(imgpath, name +
'.png') # img_fullname='/.../P000?.png'
img = Image.open(img_fullname)
img_w, img_h = img.size
# print img_w,img_h
with open(os.path.join(dstpath, name + '.txt'), 'w') as f_out:
for obj in objects:
poly = obj['poly'] # poly=[(x1,y1),(x2,y2),(x3,y3),(x4,y4)]
bbox = np.array(util.dots4ToRecC(poly, img_w,
img_h)) # bbox=[x y w h]
if (sum(bbox <= 0) +
sum(bbox >= 1)) >= 1: # 若bbox中有<=0或>= 1的元素则将该box排除
continue
if (obj['name'] in extractclassname):
id = extractclassname.index(
obj['name']) # id=类名的索引 比如'plane'对应id=0
else:
continue
outline = str(id) + ' ' + ' '.join(list(map(
str, bbox))) # outline='id x y w h'
f_out.write(outline + '\n') # 写入txt文件中并加上换行符号 \n
def HRSC2COCOTest(srcpath, destfile, cls_names):
imageparent = os.path.join(srcpath, 'images')
# labelparent = os.path.join(srcpath, 'labelTxt')
data_dict = {}
info = {'contributor': 'Jian Ding',
'data_created': '2019',
'description': 'This is HRSC.',
'url': 'http://captain.whu.edu.cn/DOTAweb/',
'version': '1.0',
'year': 2018}
data_dict['info'] = info
data_dict['images'] = []
data_dict['categories'] = []
for idex, name in enumerate(cls_names):
single_cat = {'id': idex + 1, 'name': name, 'supercategory': name}
data_dict['categories'].append(single_cat)
inst_count = 1
image_id = 1
with open(destfile, 'w') as f_out:
filenames = util.GetFileFromThisRootDir(imageparent)
# with open(test_set_file, 'r') as f_in:
# lines = f_in.readlines()
# filenames = [os.path.join(imageparent, x.strip()) + '.bmp' for x in lines]
for file in filenames:
basename = util.custombasename(file)
# image_id = int(basename[1:])
imagepath = os.path.join(imageparent, basename + '.bmp')
# img = cv2.imread(imagepath)
img = Image.open(imagepath)
# height, width, c = img.shape
height = img.height
width = img.width
single_image = {}
single_image['file_name'] = basename + '.bmp'
single_image['id'] = image_id
single_image['width'] = width
single_image['height'] = height
data_dict['images'].append(single_image)
image_id = image_id + 1
json.dump(data_dict, f_out)
def rotate(srcpath, dstpath):
pool = Pool(16)
imgnames = util.GetFileFromThisRootDir(os.path.join(srcpath, 'images'))
names = [util.custombasename(x) for x in imgnames]
dst_imgpath = os.path.join(dstpath, 'images')
dst_labelTxt = os.path.join(dstpath, 'labelTxt')
if not os.path.exists(dst_imgpath):
os.mkdir(dst_imgpath)
if not os.path.exists(dst_labelTxt):
os.mkdir(dst_labelTxt)
rotate_fun = partial(rotate_single_run, srcpath=srcpath, dstpath=dstpath)
pool.map(rotate_fun, names)
def extract_largesize_index(labelpath):
filenames = util.GetFileFromThisRootDir(labelpath)
large_size_index = []
for name in filenames:
objs = util.parse_dota_poly(name)
flag = 0
for obj in objs:
poly = np.array(obj['poly'])
xmin, ymin, xmax, ymax = np.min(poly[:, 0]), np.min(poly[:, 1]), np.max(poly[:, 0]), np.max(poly[:, 1])
w = xmax - xmin
h = ymax - ymin
max_side = max(w, h)
if max_side > 400:
flag = 1
break
if flag:
large_size_index.append(util.custombasename(name))
# print('index:', large_size_index)
# print('len:', len(large_size_index))
return large_size_index
def drwaBox_parallel(srcpath, imgpath, dstpath):
pool = Pool(16)
filelist = util.GetFileFromThisRootDir(srcpath)
nameboxdict = {}
for file in filelist:
name = util.custombasename(file)
with open(file, 'r') as f_in:
lines = f_in.readlines()
splitlines = [x.strip().split(' ') for x in lines]
for splitline in splitlines:
oriname = splitline[0]
confidence = float(splitline[1])
bbox = list(map(float, splitline[2:]))
bbox.append(confidence)
bbox.append(name.split('_')[-1])
if (oriname not in nameboxdict):
nameboxdict[oriname] = []
nameboxdict[oriname].append(bbox)
drawBoxsingle_fn = partial(drawBoxsingle, imgpath, dstpath, nameboxdict)
# pdb.set_trace()
pool.map(drawBoxsingle_fn, nameboxdict)
def mergebase(srcpath, dstpath, nms, isTask2=True):
filelist = util.GetFileFromThisRootDir(srcpath)
for fullname in filelist:
name = util.custombasename(fullname)
#print('name:', name)
dstname = os.path.join(dstpath, name + '.txt')
with open(fullname, 'r') as f_in:
nameboxdict = {}
lines = f_in.readlines()
splitlines = [x.strip().split(' ') for x in lines]
for splitline in splitlines:
subname = splitline[0]
'''
splitname = subname.split('__')
oriname = splitname[0]
pattern1 = re.compile(r'__\d+___\d+')
#print('subname:', subname)
x_y = re.findall(pattern1, subname)
x_y_2 = re.findall(r'\d+', x_y[0])
x, y = int(x_y_2[0]), int(x_y_2[1])
pattern2 = re.compile(r'__([\d+\.]+)__\d+___')
rate = re.findall(pattern2, subname)[0]
'''
splitname = subname.split('_')
oriname = splitname[0]
x = int(splitname[2])
y = int(splitname[3])
rate = int(splitname[1])
confidence = splitline[1]
poly = list(map(float, splitline[2:]))
if isTask2:
x1 = np.round(poly[0])
y1 = np.round(poly[1])
x2 = np.round(poly[0] + poly[2] - 1)
y2 = np.round(poly[1] + poly[3] - 1)
poly = [x1, y1, x2, y1, x2, y2, x1, y2]
origpoly = poly2origpoly(poly, x, y, rate)
det = origpoly
det.append(confidence)
det = list(map(float, det))
if (oriname not in nameboxdict):
nameboxdict[oriname] = []
nameboxdict[oriname].append(det)
nameboxnmsdict = nmsbynamedict(nameboxdict, nms, nms_thresh)
with open(dstname, 'w') as f_out:
for imgname in nameboxnmsdict:
for det in nameboxnmsdict[imgname]:
#print('det:', det)
confidence = det[-1]
bbox = det[0:-1]
if not isTask2:
outline = imgname + ' ' + str(
confidence) + ' ' + ' '.join(map(str, bbox))
else:
outline = imgname + ' ' + str(
confidence) + ' ' + ' '.join(
map(str,
[bbox[0], bbox[1], bbox[4], bbox[5]]))
#print('outline:', outline)
f_out.write(outline + '\n')
def dota2LongSideFormat(imgpath, txtpath, dstpath, extractclassname):
"""
trans dota farmat to longside format
:param imgpath: the path of images
:param txtpath: the path of txt in dota format
:param dstpath: the path of txt in YOLO format
:param extractclassname: the category you selected
"""
if os.path.exists(dstpath):
shutil.rmtree(dstpath) # delete output folder
os.makedirs(dstpath) # make new output folder
filelist = util.GetFileFromThisRootDir(
txtpath) # fileist=['/.../P0005.txt', ..., /.../P000?.txt]
for fullname in filelist: # fullname='/.../P000?.txt'
objects = util.parse_dota_poly(fullname)
'''
objects =
[{'name': 'ship',
'difficult': '1',
'poly': [(1054.0, 1028.0), (1063.0, 1011.0), (1111.0, 1040.0), (1112.0, 1062.0)],
'area': 1159.5
},
...
]
'''
name = os.path.splitext(os.path.basename(fullname))[0] # name='P000?'
img_fullname = os.path.join(imgpath, name +
'.png') # img_fullname='/.../P000?.png'
img = Image.open(img_fullname)
img_w, img_h = img.size
# print img_w,img_h
with open(os.path.join(dstpath, name + '.txt'), 'w') as f_out:
num_gt = 0
for i, obj in enumerate(objects):
num_gt = num_gt + 1 # 为当前有效gt计数
poly = obj['poly'] # poly=[(x1,y1),(x2,y2),(x3,y3),(x4,y4)]
poly = np.float32(np.array(poly))
# 四点坐标归一化
poly[:, 0] = poly[:, 0] / img_w
poly[:, 1] = poly[:, 1] / img_h
rect = cv2.minAreaRect(poly) # 得到最小外接矩形的（中心(x,y), (宽,高), 旋转角度）
# box = np.float32(cv2.boxPoints(rect)) # 返回rect四个点的值
c_x = rect[0][0]
c_y = rect[0][1]
w = rect[1][0]
h = rect[1][1]
theta = rect[-1] # Range for angle is [-90，0)
trans_data = cvminAreaRect2longsideformat(
c_x, c_y, w, h, theta)
if not trans_data:
if theta != 90: # Θ=90说明wh中有为0的元素，即gt信息不完整，无需提示异常，直接删除
print('opencv表示法转长边表示法出现异常,已将第%d个box排除,问题出现在该图片中:%s' %
(i, img_fullname))
num_gt = num_gt - 1
continue
else:
# range:[-180，0)
c_x, c_y, longside, shortside, theta_longside = trans_data
bbox = np.array((c_x, c_y, longside, shortside))
if (sum(bbox <= 0) +
sum(bbox[:2] >= 1)) >= 1: # 0<xy<1, 0<side<=1
print(
'bbox[:2]中有>= 1的元素,bbox中有<= 0的元素,已将第%d个box排除,问题出现在该图片中:%s'
% (i, img_fullname))
print(
'出问题的longside形式数据:[%.16f, %.16f, %.16f, %.16f, %.1f]' %
(c_x, c_y, longside, shortside, theta_longside))
num_gt = num_gt - 1
continue
if (obj['name'] in extractclassname):
id = extractclassname.index(
obj['name']) # id=类名的索引 比如'plane'对应id=0
else:
print('预定类别中没有类别:%s;已将该box排除,问题出现在该图片中:%s' %
(obj['name'], fullname))
num_gt = num_gt - 1
continue
theta_label = int(theta_longside +
180.5) # range int[0,180] 四舍五入
if theta_label == 180: # range int[0,179]
theta_label = 179
# outline='id x y longside shortside Θ'
# final check
if id > 15 or id < 0:
print('id problems,问题出现在该图片中:%s' % (i, img_fullname))
print(
'出问题的longside形式数据:[%.16f, %.16f, %.16f, %.16f, %.1f]' %
(c_x, c_y, longside, shortside, theta_longside))
if theta_label < 0 or theta_label > 179:
print('id problems,问题出现在该图片中:%s' % (i, img_fullname))
print(
'出问题的longside形式数据:[%.16f, %.16f, %.16f, %.16f, %.1f]' %
(c_x, c_y, longside, shortside, theta_longside))
outline = str(id) + ' ' + ' '.join(list(map(
str, bbox))) + ' ' + str(theta_label)
f_out.write(outline + '\n') # 写入txt文件中并加上换行符号 \n
if num_gt == 0:
os.remove(os.path.join(dstpath, name + '.txt')) #
os.remove(img_fullname)
os.remove(fullname)
print('%s 图片对应的txt不存在有效目标,已删除对应图片与txt' % img_fullname)
print('已完成文件夹内DOTA数据形式到长边表示法的转换')
import dota_utils as util
except:
import dota_kit.dota_utils as util
srcpath1 = '/mnt/lustre/yanhongchang/project/one-rpn/mmdetection/demo/work_dirs/out_img/faster_x101/crop512/txt2'
srcpath2 = '/mnt/lustre/yanhongchang/project/one-rpn/mmdetection/demo/work_dirs/out_img/faster_x101/crop1024/txt2'
srcpath3 = '/mnt/lustre/yanhongchang/project/one-rpn/mmdetection/demo/work_dirs/out_img/faster_x101/origin/txt2'
dstpath = '/mnt/lustre/yanhongchang/project/one-rpn/mmdetection/demo/work_dirs/out_img/faster_x101/tr_merge3/txt'
def osp(savepath):
if not os.path.exists(savepath):
os.makedirs(savepath)
osp(dstpath)
filelist = util.GetFileFromThisRootDir(srcpath1)
for filename in filelist:
name = util.custombasename(filename)
print(name)
file_1 = open(os.path.join(srcpath1, name + '.txt'), 'r')
file_2 = open(os.path.join(srcpath2, name + '.txt'), 'r')
file_3 = open(os.path.join(srcpath3, name + '.txt'), 'r')
file_new = open(os.path.join(dstpath, name + '.txt'), 'w')
list1 = []
for line in file_1.readlines():
ss = line.strip()
list1.append(ss)
file_1.close()
list2 = []
def prepare():
args = parse_args()
data_root_path = args.data_path
train_path = os.path.join(data_root_path, 'train')
val_path = os.path.join(data_root_path, 'val')
test_path = os.path.join(data_root_path, 'test')
if not os.path.exists(os.path.join(data_root_path, 'trainval_large')):
os.makedirs(os.path.join(data_root_path, 'trainval_large'))
if not os.path.exists(os.path.join(data_root_path, 'trainval_large', 'images')):
os.makedirs(os.path.join(data_root_path, 'trainval_large', 'images'))
if not os.path.exists(os.path.join(data_root_path, 'trainval_large', 'labelTxt')):
os.makedirs(os.path.join(data_root_path, 'trainval_large', 'labelTxt'))
if not os.path.exists(os.path.join(data_root_path, 'trainval1024_1')):
os.makedirs(os.path.join(data_root_path, 'trainval1024_1'))
split_train = ImgSplit_multi_process.splitbase(train_path,
os.path.join(data_root_path, 'trainval1024_1'),
gap=200,
subsize=1024,
num_process=args.num_process
)
split_train.splitdata(1)
split_val = ImgSplit_multi_process.splitbase(val_path,
os.path.join(data_root_path, 'trainval1024_1'),
gap=200,
subsize=1024,
num_process=args.num_process
)
split_val.splitdata(1)
# extract train images contain large intances
train_large_names = extract_largesize_index(os.path.join(data_root_path, 'train', 'labelTxt'))
filecopy(os.path.join(data_root_path, 'train', 'labelTxt'),
os.path.join(data_root_path, 'trainval_large', 'labelTxt'),
train_large_names,
'.txt',
num_process=args.num_process)
filecopy(os.path.join(data_root_path, 'train', 'images'),
os.path.join(data_root_path, 'trainval_large', 'images'),
train_large_names,
'.png',
num_process=args.num_process)
# extract val images contain large instances
val_large_names = extract_largesize_index(os.path.join(data_root_path, 'val', 'labelTxt'))
filecopy(os.path.join(data_root_path, 'val', 'labelTxt'),
os.path.join(data_root_path, 'trainval_large', 'labelTxt'),
val_large_names,
'.txt',
num_process=args.num_process)
filecopy(os.path.join(data_root_path, 'val', 'images'),
os.path.join(data_root_path, 'trainval_large', 'images'),
val_large_names,
'.png',
num_process=args.num_process)
# split for images contin large size instances
if not os.path.exists(os.path.join(data_root_path, 'trainval_large_1024_0.4')):
os.makedirs(os.path.join(data_root_path, 'trainval_large_1024_0.4'))
split_trainval_large = ImgSplit_multi_process.splitbase(os.path.join(data_root_path, 'trainval_large'),
os.path.join(data_root_path, 'trainval_large_1024_0.4'),
gap=512,
subsize=1024,
num_process=args.num_process)
split_trainval_large.splitdata(0.4)
# rotate augment for images contain large size instances
rotate_augment(os.path.join(data_root_path, 'trainval_large_1024_0.4'),
os.path.join(data_root_path, 'trainval_large_1024_0.4_rotate'))
# copy files to images and labelTxt
if not os.path.exists(os.path.join(data_root_path, 'images')):
os.makedirs(os.path.join(data_root_path, 'images'))
if not os.path.exists(os.path.join(data_root_path, 'labelTxt')):
os.makedirs(os.path.join(data_root_path, 'labelTxt'))
filemove_v2(os.path.join(data_root_path, 'trainval1024_1', 'images'),
os.path.join(data_root_path, 'images'),
'.png',
num_process=args.num_process
)
filemove_v2(os.path.join(data_root_path, 'trainval1024_1', 'labelTxt'),
os.path.join(data_root_path, 'labelTxt'),
'.txt',
num_process=args.num_process
)
filemove_v2(os.path.join(data_root_path, 'trainval_large_1024_0.4', 'images'),
os.path.join(data_root_path, 'images'),
'.png',
num_process=args.num_process
)
filemove_v2(os.path.join(data_root_path, 'trainval_large_1024_0.4', 'labelTxt'),
os.path.join(data_root_path, 'labelTxt'),
'.txt',
num_process=args.num_process
)
filemove_v2(os.path.join(data_root_path, 'trainval_large_1024_0.4_rotate', 'images'),
os.path.join(data_root_path, 'images'),
'.png',
num_process=args.num_process
)
filemove_v2(os.path.join(data_root_path, 'trainval_large_1024_0.4_rotate', 'labelTxt'),
os.path.join(data_root_path, 'labelTxt'),
'.txt',
num_process=args.num_process
)
train_without_balance = util.GetFileFromThisRootDir(os.path.join(data_root_path, 'labelTxt'))
train_without_balance_names = [util.custombasename(x.strip()) for x in train_without_balance]
# data balance
with open('train_balance_extend.txt', 'r') as f_in:
train_balance_names = f_in.readlines()
train_balance_names = [x.strip() for x in train_balance_names]
train_names = train_without_balance_names + train_balance_names
with open(os.path.join(data_root_path, 'train.txt'), 'w') as f_out:
for index, name in enumerate(train_names):
if index == (len(train_names) - 1):
f_out.write(name)
else:
f_out.write(name + '\n')
# prepare test data
if not os.path.exists(os.path.join(data_root_path, 'test1024')):
os.makedirs(os.path.join(data_root_path, 'test1024'))
split_test = SplitOnlyImage_multi_process.splitbase(os.path.join(test_path, 'images'),
os.path.join(data_root_path, 'test1024', 'images'),
gap=512,
subsize=1024,
num_process=args.num_process
)
split_test.splitdata(1)
split_test.splitdata(0.5)
test_names = util.GetFileFromThisRootDir(os.path.join(data_root_path, 'test1024', 'images'))
test_names = [util.custombasename(x.strip()) for x in test_names]
with open(os.path.join(data_root_path, 'test.txt'), 'w') as f_out:
for index, name in enumerate(test_names):
if index == (len(test_names) - 1):
f_out.write(name)
else:
f_out.write(name + '\n')
filemove_v2(os.path.join(data_root_path, 'test1024', 'images'),
os.path.join(data_root_path, 'images'),
'.png',
num_process=args.num_process)
shutil.rmtree(os.path.join(data_root_path, r'trainval_large_1024_0.4'))
shutil.rmtree(os.path.join(data_root_path, r'trainval_large_1024_0.4_rotate'))
shutil.rmtree(os.path.join(data_root_path, r'test1024'))
shutil.rmtree(os.path.join(data_root_path, r'trainval1024_1'))
shutil.rmtree(os.path.join(data_root_path, r'trainval_large'))
def filemove_v2(srcpath, dstpath, extent, num_process=32):
filelist = util.GetFileFromThisRootDir(srcpath)
filenames = [util.custombasename(x.strip()) for x in filelist]
print('srcpath: ', srcpath)
print('num: ', len(filenames))
filemove(srcpath, dstpath, filenames, extent, num_process)
def mergebase(srcpath, dstpath, nms):
filelist = util.GetFileFromThisRootDir(srcpath)
for filename in filelist:
mergesingle(dstpath, nms, filename)