主页 / 问题 / 970467
Accepted
Егор Никоноров
Asked:2020-04-17 00:48:22 +0000 UTC2020-04-17 00:48:22 +0000 UTC

如何在数组中显示所选对象的极值点坐标？

  • 772

物体识别有一段代码：

from imageai.Detection import VideoObjectDetection
import os

execution_path = os.getcwd()

detector = VideoObjectDetection()
detector.setModelTypeAsYOLOv3()
detector.setModelPath( os.path.join(execution_path , "yolo.h5"))
detector.loadModel()

video_path = detector.detectObjectsFromVideo(
    input_file_path = os.path.join(execution_path, "Car - 2165.mp4"),
    output_file_path = os.path.join(execution_path, "traffic_detected"),
    frames_per_second = 20,
    log_progress = True
)

assert isinstance(video_path, object)
print(video_path)

如何在数组（具体来说，矩形等）中显示所选对象的极值点坐标？

python

1 个回答

  1. Best Answer

    detector.detectObjectsFromVideo()函数可以作为参数传递给函数per_frame_function。处理完每一帧后，将调用传递的函数。此函数在调用时将传递以下参数：

    • frame_number- 帧的序列号

    • output_array- 在帧中检测到的对象数组，由以下形式的字典组成：

      {'box_points': (362, 295, 443, 355), 'name': 'boat', 'percentage_probability': 26.67}

    • output_count- 统计：每个类的对象数。例子：

      {'bus': 4, 'boat': 3, 'person': 1, 'car': 8}

    为了一起收集这些数据，我会使用Pandas模块：

    import pandas as pd

data = []

def collect_detected_objs(*args):
    data.append(args)

ret = detector.detectObjectsFromVideo(
    input_file_path = video_fn,
    output_file_path = os.path.join(execution_path, "traffic_detected"),
    frames_per_second = 20,
    log_progress = True,
    per_frame_function=collect_detected_objs
)

d = pd.concat([pd.DataFrame(x[1]).assign(frame=x[0]) for x in data])
stats = pd.DataFrame([x[2] for x in data], index=[x[0] for x in data])

    结果：

    In [233]: d
Out[233]:
                 box_points    name  percentage_probability  frame
0       (504, 53, 590, 163)   truck               50.516695      1
1   (1008, 850, 1222, 1077)   truck               52.645624      1
2      (768, 124, 868, 261)   truck               75.007182      1
3      (534, 164, 600, 233)     car               63.670057      1
4   (1008, 850, 1222, 1077)     car               64.012623      1
5      (625, 121, 692, 239)     car               68.127060      1
6        (584, 24, 633, 87)     car               83.921510      1
7      (638, 413, 661, 472)  person               53.726447      1
8      (445, 722, 475, 789)  person               58.984256      1
9      (481, 673, 507, 719)  person               61.079180      1
..                      ...     ...                     ...    ...
8      (423, 873, 452, 936)  person               58.492571     47
0      (768, 125, 868, 261)   truck               68.915749     48
1      (537, 183, 597, 244)     car               78.283107     48
2   (1010, 850, 1223, 1079)     car               78.731704     48
3        (577, 13, 627, 72)     car               85.132539     48
4      (620, 156, 694, 203)     car               90.549046     48
5      (860, 458, 886, 521)  person               50.013673     48
6      (423, 873, 450, 936)  person               53.440607     48
7      (397, 745, 424, 809)  person               57.603621     48
8      (628, 396, 650, 447)  person               59.033847     48

[549 rows x 4 columns]

In [234]: stats
Out[234]:
    car  person  truck
1     4       5      3
2     4       5      3
3     4       6      3
4     4       6      2
5     4       6      3
6     4       6      4
7     4       6      3
8     4       7      3
9     4       7      3
10    4       7      3
..  ...     ...    ...
39    5       4      2
40    5       4      2
41    5       5      1
42    5       1      1
43    5       2      1
44    4       2      1
45    4       2      1
46    4       3      1
47    4       4      1
48    4       4      1

[48 rows x 3 columns]

    整个代码：

    import os
import cv2
from imageai.Detection import ObjectDetection, VideoObjectDetection
import pandas as pd

execution_path = os.getcwd()

yolo_model_fn = r'C:\work\ML\.data\misc\yolo.h5'
video_fn = r'C:\download\cut2.mp4'

detector = VideoObjectDetection()
detector.setModelTypeAsYOLOv3()
detector.setModelPath(yolo_model_fn)
detector.loadModel()

data = []

def collect_detected_objs(*args):
    data.append(args)
    #detected.append(pd.DataFrame(detections).assign(frame=frame_pos))
    #stats.append(pd.Series(obj_stats, index=[frame_pos]))

ret = detector.detectObjectsFromVideo(
    input_file_path = video_fn,
    output_file_path = os.path.join(execution_path, "traffic_detected"),
    frames_per_second = 20,
    log_progress = True,
    per_frame_function=collect_detected_objs
)

d = pd.concat([pd.DataFrame(x[1]).assign(frame=x[0]) for x in data])
stats = pd.DataFrame([x[2] for x in data], index=[x[0] for x in data])

print(d)
print(stats)

    • 2

相关问题