Tensorflow对象检测中的裁剪框，并将其显示为jpg图像

Question

我使用tensorflow异议检测来检测护照上的特定数据，如全名和其他内容。我已经对数据进行了训练，一切都很好。它用一个边界框完美地识别了它周围的数据。然而，现在我只想裁剪检测到的框。

代码：

import os
import cv2
import numpy as np
import tensorflow as tf
import sys
sys.path.append("..")
from object_detection.utils import label_map_util
from object_detection.utils import visualization_utils as vis_util
MODEL_NAME = 'inference_graph'
CWD_PATH = os.getcwd()
PATH_TO_CKPT = 'C:/Users/UI UX/Desktop/Captcha 3/CAPTCHA_frozen_inference_graph.pb'
PATH_TO_LABELS = 'C:/Users/UI UX/Desktop/Captcha 3/CAPTCHA_labelmap.pbtxt'
PATH_TO_IMAGE = 'C:/Users/UI UX/Desktop/(47).jpg'
NUM_CLASSES = 11
label_map = label_map_util.load_labelmap(PATH_TO_LABELS)
categories = label_map_util.convert_label_map_to_categories(label_map, max_num_classes=NUM_CLASSES, use_display_name=True)
category_index = label_map_util.create_category_index(categories)
detection_graph = tf.Graph()
with detection_graph.as_default():
od_graph_def = tf.GraphDef()
with tf.gfile.GFile(PATH_TO_CKPT, 'rb') as fid:
serialized_graph = fid.read()
od_graph_def.ParseFromString(serialized_graph)
tf.import_graph_def(od_graph_def, name='')
sess = tf.Session(graph=detection_graph)
image_tensor = detection_graph.get_tensor_by_name('image_tensor:0')
detection_boxes = detection_graph.get_tensor_by_name('detection_boxes:0')
detection_scores = detection_graph.get_tensor_by_name('detection_scores:0')
detection_classes = detection_graph.get_tensor_by_name('detection_classes:0')
num_detections = detection_graph.get_tensor_by_name('num_detections:0')
image = cv2.imread(PATH_TO_IMAGE)
image_np = cv2.resize(image, (0, 0), fx=2.0, fy=2.0)
image_expanded = np.expand_dims(image_np, axis=0)
(boxes, scores, classes, num) = sess.run(
[detection_boxes, detection_scores, detection_classes, num_detections],
feed_dict={image_tensor: image_expanded})
vis_util.visualize_boxes_and_labels_on_image_array(
image_np,
np.squeeze(boxes),
np.squeeze(classes).astype(np.int32),
np.squeeze(scores),
category_index,
use_normalized_coordinates=True,
line_thickness=2,
min_score_thresh=0.60)
width, height = image_np.shape[:2]
for i, box in enumerate(np.squeeze(boxes)):
if(np.squeeze(scores)[i] > 0.80):
(ymin, xmin, ymax, xmax) = (box[0]*height, box[1]*width, box[2]*height, box[3]*width)
cropped_image = tf.image.crop_to_bounding_box(image_np, ymin, xmin, ymax - ymin, xmax - xmin)
cv2.imshow('cropped_image', image_np)
cv2.waitKey(0)
cv2.imshow('Object detector', image_np)
cv2.waitKey(0)
cv2.destroyAllWindows()

但是得到这个错误：

追踪(最近一次通话)：文件"；C：/Users/UI-UX/PycharmProjects/pythonProject1/vedio_object_detection.py"；，第71行，incropped_image=tf.image.crop_to_bounding_box(image_np，ymin，xmin，ymax-ymin，xmax-xmin)文件"；C： \ProgramData\Anaconda2\envs\tf_cpu\lib\site packages\tensorflow_core\python\ops\image_ops_impl.py"，第875行，在crop_to_bounding_box中array_ops.stack([-1，target_height，target_width，-1])文件"；C： \ProgramData\Anaconda2\envs\tf_cpu\lib\site packages\tensorflow_core\python\ops\array_ops.py"，第855行，切片返回gen_array_ops切片(输入，begin，size，name=name)文件"；C： \ProgramData\Anaconda2\envs\tf_cpu\lib\site packages\tensorflow_core\python\ops\gen_array_ops.py"，第9222行，在_slice中"切片"；，input=input，begin=begin，size=size，name=name)文件"；C： \ProgramData\Anaconda2\envs\tf_cpu\lib\site packages\tensorflow_core\python\framework\op_def_library.py"，第632行，在_apply_op_helper中param_name=输入名称)文件"；C： \ProgramData\Anaconda2\envs\tf_cpu\lib\site packages\tensorflow_core\python\framework\op_def_library.py"，第61行，在_SompatiiesTypeConstraint中&"&"；。join(dtypes.as_dtype(x).allowed_list中x的名称))TypeError:传递给参数"begin"的值的DataType float32不在允许的值列表中：int32，int64

有什么帮助吗？

Answer 1

我通过在这一行的末尾添加以下代码找到了解决方案：

(boxes, scores, classes, num) = sess.run([detection_boxes, detection_scores, detection_classes, num_detections],feed_dict={image_tensor: image_expanded})

我加上这个：

(frame_height, frame_width) = image.shape[:2]
for i in range(len(np.squeeze(scores))):
#print(np.squeeze(boxes)[i])
ymin = int((np.squeeze(boxes)[i][0]*frame_height))
xmin = int((np.squeeze(boxes)[i][1]*frame_width))
ymax = int((np.squeeze(boxes)[i][2]*frame_height))
xmax = int((np.squeeze(boxes)[i][3]*frame_width))
cropped_img = image[ymax:ymin,xmax:xmin]
cv2.imwrite(f'/your/path/img_{i}.png', cropped_img)

Answer 2

对于寻找人体检测边界框解决方案的人来说，这里有一个快速的方法。当你进行人体检测时，它会给你一个边界框，x1，y1是opencv坐标，x2和y2是宽度和高度。如果你想在人体检测后绘制矩形或裁剪图像，你可以使用以下代码

top = box[0]
left = box[1]
bottom = top+box[2]
right = left+box[3]
cv2.rectangle(image, (top,left),(bottom,right), (0, 255, 0), 2)