def __call__(self, in_data):
img, label = in_data
_, H, W = img.shape
img = preprocess(img, self.min_size, self.max_size)
_, o_H, o_W = img.shape
scale = o_H / H
# horizontally flip
img, params = util.random_flip(img, x_random=True, return_param=True)
return img, label, scale
def __call__(self, in_data):
img, bbox, label = in_data
_, H, W = img.shape
img = preprocess(img, self.min_size, self.max_size)#对图像缩放处理
_, o_H, o_W = img.shape
scale = o_H / H
bbox = util.resize_bbox(bbox, (H, W), (o_H, o_W))#对bbox处理
#图像随机水平翻转
img, params = util.random_flip(img, x_random=True, return_param=True)
#对应的bbox
bbox = util.flip_bbox(bbox, (o_H, o_W), x_flip=params['x_flip'])
return img, bbox, label, scale
def __call__(self, in_data):
img, bbox, label = in_data
_, H, W = img.shape
img = preprocess(img, self.min_size, self.max_size)
_, o_H, o_W = img.shape
scale = o_H / H
bbox = util.resize_bbox(bbox, (H, W), (o_H, o_W))
# horizontally flip
img, params = util.random_flip(img, x_random=True, return_param=True)
bbox = util.flip_bbox(bbox, (o_H, o_W), x_flip=params['x_flip'])
return img, bbox, label, scale
def __call__(self, in_data):
img, bbox, label = in_data
_, H, W = img.shape
img = preprocess(img, self.min_size, self.max_size)
_, o_H, o_W = img.shape # 缩放后的图像大小
scale = o_H / H # 缩放比例
bbox = util.resize_bbox(bbox, (H, W), (o_H, o_W)) # 按比例缩放包围盒
# horizontally flip， 将图像随机水平翻转
img, params = util.random_flip(img, x_random=True, return_param=True)
# 将包围盒进行与图像相同的水平翻转
bbox = util.flip_bbox(bbox, (o_H, o_W), x_flip=params['x_flip'])
return img, bbox, label, scale
def __call__(self, data):
img, bbox, label = data
_, H, W = img.shape
img = preprocess(img)
_, o_H, o_W = img.shape
bbox = resize_bbox(bbox, (H, W), (o_H, o_W))
#randomly horizontally flip
img, x_flip = random_flip(img)
if x_flip:
bbox = flip_bbox(bbox, (o_H, o_W))
scale = o_H / H
return img, bbox, label, scale
def __call__(self, in_data):
img, bbox, label = in_data
_, H, W = img.shape
img = preprocess(img, self.min_size, self.max_size) #预处理图片
_, o_H, o_W = img.shape
scale = o_H / H
#预处理图片中的bbox,对相应的bounding boxes 也进行同等尺度的缩放。
bbox = util.resize_bbox(bbox, (H, W), (o_H, o_W))
# horizontally flip水平翻转
img, params = util.random_flip(img, x_random=True, return_param=True)
bbox = util.flip_bbox(bbox, (o_H, o_W), x_flip=params['x_flip'])
return img, bbox, label, scale
def __call__(self, in_data):
img, bbox, label = in_data
_, H, W = img.shape
img = preprocess(img, self.min_size,
self.max_size) # 将图片进行最小最大化放缩然后进行归一化
_, o_H, o_W = img.shape
scale = o_H / H # 放缩前后相除，得出放缩比因子
bbox = util.resize_bbox(bbox, (H, W), (o_H, o_W)) # 重新调整bboxes框的大小
# 水平翻转
# 进行图片的随机反转，图片旋转不变性，增强网络的鲁棒性！
img, params = util.random_flip(img, x_random=True, return_param=True)
# 同样的对bboxes进行随机反转
bbox = util.flip_bbox(bbox, (o_H, o_W), x_flip=params['x_flip'])
return img, bbox, label, scale
def __call__(self, in_data):
img, bbox, label = in_data
_, H, W = img.shape #(3,h,w)
img = preprocess(img, self.min_size,
self.max_size) #(3,h*scale,w*scale)
_, o_H, o_W = img.shape
scale = o_H / H #(得出缩放比因子) #scale
bbox = util.resize_bbox(bbox, (H, W), (o_H, o_W)) #
# horizontally flip
img, params = util.random_flip(
img, x_random=True,
return_param=True) #jiang pictures shuiping fanzhuang
bbox = util.flip_bbox(bbox, (o_H, o_W), x_flip=params['x_flip'])
return img, bbox, label, scale
def __call__(self, in_data):
img, bbox, label = in_data
_, H, W = img.shape
# 图像等比例缩放
img = preprocess(img, self.min_size, self.max_size)
_, o_H, o_W = img.shape
# 得出缩放比因子
scale = o_H / H
# bbox按照与原图等比例缩放
bbox = util.resize_bbox(bbox, (H, W), (o_H, o_W))
# 将图片进行随机水平翻转，没有进行垂直翻转
img, params = util.random_flip(
img, x_random=True, return_param=True)
# 同样地将bbox进行与对应图片同样的水平翻转
bbox = util.flip_bbox(
bbox, (o_H, o_W), x_flip=params['x_flip'])
return img, bbox, label, scale
def __call__(self, in_data):
img, bbox, label = in_data
_, H, W = img.shape
# print("before preprocessing:", img.shape, img[:, 100: 200, 100: 200])
img = preprocess(img, self.min_size, self.max_size)
# print("after preprocessing:", img.shape, img[:, 100: 200, 100: 200])
_, o_H, o_W = img.shape
# TODO change the defenition of scale.
# warning: might have issue.
scale = o_H / H
# print("bbox pre: ", bbox)
bbox = util.resize_bbox(bbox, (H, W), (o_H, o_W))
# print("bbox after: ", bbox)
# horizontally flip
img, params = util.random_flip(img, x_random=True, return_param=True)
bbox = util.flip_bbox(bbox, (o_H, o_W), x_flip=params['x_flip'])
# print("bbox after flip: ", bbox, bbox.shape)
return img, bbox, label, scale
def __call__(self, in_data):
img, bbox, label, depth, y_rot = in_data
_, H, W = img.shape
img = preprocess(img, self.min_size, self.max_size)
_, o_H, o_W = img.shape
scale = o_H / H
bbox = util.resize_bbox(bbox, (H, W), (o_H, o_W))
# horizontally flip
img, params = util.random_flip(img, x_random=True, return_param=True)
bbox = util.flip_bbox(bbox, (o_H, o_W), x_flip=params['x_flip'])
if params['x_flip']:
for i in range(len(y_rot)):
theta = float(y_rot[i])
if theta > 0:
y_rot[i] = pi - theta
if theta < 0:
y_rot[i] = -pi - theta
y_rot[i]
if theta == 0:
y_rot[i] = 3.14
return img, bbox, label, depth, y_rot, scale
