def trim(im, *, border=0, color=None):
""" 默认裁剪掉白色边缘，可以配合 get_backgroup_color 裁剪掉背景色
:param border: 上下左右保留多少边缘
输入一个整数，表示统一留边
也可以输入[int, int, int, int]，分别表示left top right bottom留白
:param color: 要裁剪的颜色，这是精确值，没有误差，如果需要模糊，可以提前预处理成精确值
这个默认值设成None，本来是想说支持灰度图，此时输入一个255的值就好
但是pil的灰度图机制好像有些不太一样，总之目前默认值还是自动会设成 (255, 255, 255)
:param percent-background: TODO 控制裁剪百分比上限
"""
from PIL import Image, ImageChops
from pyxllib.algo.geo import xywh2ltrb, ltrb2xywh, ltrb_border
if color is None:
color = (255, 255, 255)
bg = Image.new(im.mode, im.size, color)
diff = ImageChops.difference(im, bg)
bbox = diff.getbbox() # 如果im跟bg一样，也就是裁"消失"了，此时bbox值为None
if bbox:
if border:
ltrb = xywh2ltrb(bbox)
ltrb = ltrb_border(ltrb, border, im.size)
bbox = ltrb2xywh(ltrb)
im = im.crop(bbox)
return im
def to_icdar_label_quad(self, outfile, *, min_score=0):
""" 将coco的dt结果转为icdar的标注格式
存成一个zip文件，zip里面每张图对应一个txt标注文件
每个txt文件用quad八个数值代表一个标注框
适用于 sroie 检测格式
"""
# 1 获取dt_list
if min_score:
dt_list = [b for b in self.dt_list if (b['score'] >= min_score)]
else:
dt_list = self.dt_list
# 2 转df，按图片分组处理
df = pd.DataFrame.from_dict(dt_list) # noqa from_dict可以传入List[Dict]
df = df.groupby('image_id')
# 3 建立一个zip文件
myzip = ZipFile(str(outfile), 'w')
# 4 遍历每一组数据，生成一个文件放到zip里面
id2name = {
im['id']: pathlib.Path(im['file_name']).stem
for im in self.gt_dict['images']
}
for image_id, items in df:
label_file = id2name[image_id] + '.txt'
quads = [
rect2polygon(xywh2ltrb(x), dtype=int).reshape(-1)
for x in items['bbox']
]
quads = [','.join(map(str, x)) for x in quads]
myzip.writestr(label_file, '\n'.join(quads))
myzip.close()
def img2rects(self,
img,
haystack=None,
*,
grayscale=None,
confidence=None):
""" 根据预存的img数据，匹配出多个内容对应的所在的rect位置 """
# 1 配置参数
if isinstance(img, str):
img = self[img]['img']
grayscale = first_nonnone([grayscale, self.grayscale, False])
confidence = first_nonnone([confidence, self.confidence, 0.95])
# 2 查找子图
if haystack is None:
self.update_shot()
haystack = self.last_shot
boxes = pyautogui.locateAll(img,
haystack,
grayscale=grayscale,
confidence=confidence)
# 3 过滤掉重叠超过一半面积的框
return ComputeIou.nms_ltrb([xywh2ltrb(box) for box in list(boxes)])
def check_img(self,
loclabel,
needle=None,
*,
grayscale=None,
confidence=None):
grayscale = first_nonnone([grayscale, self.grayscale, False])
confidence = first_nonnone([confidence, self.confidence, 0.95])
if needle is None:
needle = self[loclabel]['img']
elif isinstance(needle, str):
needle = self[needle]['img']
haystack = self.screenshot(loclabel)
if confidence >= 1:
return np.array_equal(needle, haystack)
else:
boxes = pyautogui.locateAll(needle,
self.screenshot(loclabel),
grayscale=grayscale,
confidence=confidence)
boxes = [xywh2ltrb(box) for box in list(boxes)]
return len(boxes)
def bbox2ltrb(cls, b):
return [int(round(v, 0)) for v in xywh2ltrb(b)]
def to_labelme_cls(self, root, *, bbox=True, seg=False, info=False):
"""
:param root: 图片根目录
:return:
extdata，存储了一些匹配异常信息
"""
root, data = Dir(root), {}
catid2name = {x['id']: x['name'] for x in self.gt_dict['categories']}
# 1 准备工作，构建文件名索引字典
gs = PathGroups.groupby(root.select_files('**/*'))
# 2 遍历生成labelme数据
not_finds = set() # coco里有的图片，root里没有找到
multimatch = dict() # coco里的某张图片，在root找到多个匹配文件
for img, anns in tqdm(self.group_gt(reserve_empty=True),
disable=not info):
# 2.1 文件匹配
imfiles = gs.find_files(img['file_name'])
if not imfiles: # 没有匹配图片的，不处理
not_finds.add(img['file_name'])
continue
elif len(imfiles) > 1:
multimatch[img['file_name']] = imfiles
imfile = imfiles[0]
else:
imfile = imfiles[0]
# 2.2 数据内容转换
lmdict = LabelmeDict.gen_data(imfile)
img = DictTool.or_(img, {'xltype': 'image'})
lmdict['shapes'].append(
LabelmeDict.gen_shape(json.dumps(img, ensure_ascii=False),
[[-10, 0], [-5, 0]]))
for ann in anns:
if bbox:
ann = DictTool.or_(
ann, {'category_name': catid2name[ann['category_id']]})
label = json.dumps(ann, ensure_ascii=False)
shape = LabelmeDict.gen_shape(label,
xywh2ltrb(ann['bbox']))
lmdict['shapes'].append(shape)
if seg:
# 把分割也显示出来（用灰色）
for x in ann['segmentation']:
an = {
'box_id': ann['id'],
'xltype': 'seg',
'shape_color': [191, 191, 191]
}
label = json.dumps(an, ensure_ascii=False)
lmdict['shapes'].append(LabelmeDict.gen_shape(
label, x))
f = imfile.with_suffix('.json')
data[f.relpath(root)] = lmdict
return LabelmeDataset(root,
data,
extdata={
'categories': self.gt_dict['categories'],
'not_finds': not_finds,
'multimatch': Groups(multimatch)
})