使用 Tensorflow Api 和 Opencv 在视频上裁剪检测到的对象
Posted
技术标签:
【中文标题】使用 Tensorflow Api 和 Opencv 在视频上裁剪检测到的对象【英文标题】:Cropping A Detected Object On A Video With Tensorflow Api And Opencv 【发布时间】:2018-12-29 15:31:21 【问题描述】:-Python 3.6 -Tensorflow 1.11 支持 GPU。 -Opencv 3.4.2
我正在研究 Tensorflow Api,并且我已经训练了我的数据集。它工作正常。但是我必须裁剪检测到的对象并对其进行一些预处理。这看起来很简单,因为 Tensroflow 也用绿色框绘制了检测到的对象。当我尝试找到对象的坐标时,它给了我 0 到 1 范围内的数字。当我将坐标放在 Opencv Crop Image 上时,我必须将图像与图片的高度和宽度相乘,但它会出错。
Tensorflow.org 说我可以使用“tf.image.crop_and_resize”函数。但我不能在我自己的代码上运行它。
这是我的 run_inference_for_single_image 函数并返回 output_dict:
def run_inference_for_single_image(image, graph):
with graph.as_default():
#with tf.Session() as sess:
# Get handles to input and output tensors
ops = tf.get_default_graph().get_operations()
all_tensor_names = output.name for op in ops for output in op.outputs
tensor_dict =
for key in [
'num_detections', 'detection_boxes', 'detection_scores',
'detection_classes', 'detection_masks'
]:
tensor_name = key + ':0'
if tensor_name in all_tensor_names:
tensor_dict[key] = tf.get_default_graph().get_tensor_by_name(
tensor_name)
if 'detection_masks' in tensor_dict:
# The following processing is only for single image
detection_boxes = tf.squeeze(tensor_dict['detection_boxes'], [0])
detection_masks = tf.squeeze(tensor_dict['detection_masks'], [0])
# Reframe is required to translate mask from box coordinates to image coordinates and fit the image size.
real_num_detection = tf.cast(tensor_dict['num_detections'][0], tf.int32)
detection_boxes = tf.slice(detection_boxes, [0, 0], [real_num_detection, -1])
detection_masks = tf.slice(detection_masks, [0, 0, 0], [real_num_detection, -1, -1])
detection_masks_reframed = utils_ops.reframe_box_masks_to_image_masks(
detection_masks, detection_boxes, image.shape[0], image.shape[1])
detection_masks_reframed = tf.cast(
tf.greater(detection_masks_reframed, 0.5), tf.uint8)
# Follow the convention by adding back the batch dimension
tensor_dict['detection_masks'] = tf.expand_dims(
detection_masks_reframed, 0)
image_tensor = tf.get_default_graph().get_tensor_by_name('image_tensor:0')
# Run inference
output_dict = sess.run(tensor_dict,
feed_dict=image_tensor: np.expand_dims(image, 0))
# all outputs are float32 numpy arrays, so convert types as appropriate
output_dict['num_detections'] = int(output_dict['num_detections'][0])
output_dict['detection_classes'] = output_dict[
'detection_classes'][0].astype(np.uint8)
output_dict['detection_boxes'] = output_dict['detection_boxes'][0]
output_dict['detection_scores'] = output_dict['detection_scores'][0]
if 'detection_masks' in output_dict:
output_dict['detection_masks'] = output_dict['detection_masks'][0]
return output_dict
这是我的视频捕捉功能。它裁剪了错误的坐标。
video = cv2.VideoCapture(0)
ret = video.set(3,1080)
ret = video.set(4,720)
while(True):
# Acquire frame and expand frame dimensions to have shape: [1, None, None, 3]
# i.e. a single-column array, where each item in the column has the pixel RGB value
ret,frame = video.read()
frame = cv2.flip(frame, 1)
frame_expanded = np.expand_dims(frame, axis=0)
# Perform the actual detection by running the model with the image as input
(boxes, scores, classes, num) = sess.run(
[detection_boxes, detection_scores, detection_classes, num_detections],
feed_dict=image_tensor: frame_expanded)
vis_util.visualize_boxes_and_labels_on_image_array(
frame,
np.squeeze(boxes),
np.squeeze(classes).astype(np.int32),
np.squeeze(scores),
category_index,
use_normalized_coordinates=True,
line_thickness=8,
min_score_thresh=0.50)
# Draw the results of the detection (aka 'visulaize the results')
output_dict = run_inference_for_single_image(frame, detection_graph)
max_boxes_to_draw = output_dict['detection_boxes'].shape[0]
for i in range(min(max_boxes_to_draw, output_dict['detection_boxes'].shape[0])):
if output_dict['detection_scores'][i] > 0.95:
if output_dict['detection_classes'][i] in category_index.keys():
class_name = category_index[output_dict['detection_classes'][i]]['name']
print(output_dict['detection_boxes'][i])
crop_img = frame[int((output_dict['detection_boxes'][i][0]) * 720): int(
(output_dict['detection_boxes'][i][2]) * 720),
int((output_dict['detection_boxes'][i][1]) * 1080):int(
(output_dict['detection_boxes'][i][3]) * 1080)]
cv2.imshow("asdasd", crop_img)
print(class_name)
cv2.imshow('Object detector', frame)
# Press 'q' to quit
if cv2.waitKey(1) == ord('q'):
break
可能是关于 output_dict。
class_name = category_index[output_dict['detection_classes'][i]]['name'] => 这些代码给了我类的名称。效果很好。
【问题讨论】:
SO 允许将图像复制粘贴到您的问题中。查看一些输出可能会有所帮助。另外,你能验证 output_dict['detection_boxes'][i] 的格式是 [y0,x0,y1,x1] 吗?请注意,在 numpy 形式中,图像是 y-x,而不是 x-y。如果你把它弄混了,它可能解释一下。但什么也没看到,很难说。 【参考方案1】:我找到了我的问题的答案。这是解决方案代码:
while(True):
# Acquire frame and expand frame dimensions to have shape: [1, None, None, 3]
# i.e. a single-column array, where each item in the column has the pixel RGB value
ret,frame = video.read()
frame = cv2.flip(frame, 1)
frame_expanded = np.expand_dims(frame, axis=0)
# Perform the actual detection by running the model with the image as input
(boxes, scores, classes, num) = sess.run(
[detection_boxes, detection_scores, detection_classes, num_detections],
feed_dict=image_tensor: frame_expanded)
vis_util.visualize_boxes_and_labels_on_image_array(
frame,
np.squeeze(boxes),
np.squeeze(classes).astype(np.int32),
np.squeeze(scores),
category_index,
use_normalized_coordinates=True,
line_thickness=8,
min_score_thresh=0.50)
# Draw the results of the detection (aka 'visulaize the results')
output_dict = run_inference_for_single_image(frame, detection_graph)
max_boxes_to_draw = output_dict['detection_boxes'].shape[0]
for i in range(min(max_boxes_to_draw, output_dict['detection_boxes'].shape[0])):
if output_dict['detection_scores'][i] > 0.80:
if output_dict['detection_classes'][i] in category_index.keys():
class_name = category_index[output_dict['detection_classes'][i]]['name']
#print(output_dict['detection_boxes'][i])
ymin = boxes[0, i, 0]
xmin = boxes[0, i, 1]
ymax = boxes[0, i, 2]
xmax = boxes[0, i, 3]
im_width = 1280
im_height = 720
(xminn, xmaxx, yminn, ymaxx) = (xmin * im_width, xmax * im_width, ymin * im_height, ymax * im_height)
crop_img=tf.image.crop_to_bounding_box(frame,int(yminn), int(xminn), int(ymaxx-yminn), int(xmaxx-xminn))
crop_img=frame[int(yminn):int(ymaxx),int(xminn):int(xmaxx)]
# print(session.run(file))
"""crop_img = frame[int((output_dict['detection_boxes'][i][0]) * 720): int(
(output_dict['detection_boxes'][i][2]) * 720),
int((output_dict['detection_boxes'][i][1]) * 1080):int(
(output_dict['detection_boxes'][i][3]) * 1080)]"""
cv2.imshow("asdsda",crop_img)
#print(class_name)
cv2.imshow('Object detector', frame)
# Press 'q' to quit
if cv2.waitKey(1) == ord('q'):
break
im_width = 1280 对我来说没有任何意义,但它适用于我的项目,但它有效。感谢您的帮助。
【讨论】:
以上是关于使用 Tensorflow Api 和 Opencv 在视频上裁剪检测到的对象的主要内容,如果未能解决你的问题,请参考以下文章
干货使用TensorFlow官方Java API调用TensorFlow模型(附代码)
在 TensorFlow 对象检测 API 中打印类名和分数
Tensorflow 2.0—— 跟着官方demo学习Tensorflow 2.0框架基本API和用法