如何将 tensorflow 对象检测 api 与质心跟踪集成

Question

我正在使用 TensorFlow 对象检测 API 来检测视频帧中的人物。我想将检测坐标提供给质心跟踪器，为每个检测到的对象分配一个 ID，并避免在每一帧都将它们检测为新对象。

这是我用于检测和获取边界框的代码

# Import packages
import os
import cv2
import numpy as np
import tensorflow as tf
import sys

# This is needed since the notebook is stored in the object_detection folder.
from pandas._libs import json

sys.path.append("..")

# Import utilites
from utils import label_map_util
from utils import visualization_utils as vis_util

# Name of the directory containing the object detection module we're using
MODEL_NAME = 'inference_graph'
VIDEO_NAME = 'test.mp4'

# direction for out put file
FILE_OUTPUT = 'C:/Mohammad/tensorflow1/models/research/object_detection/savedframes/out.avi'

# Grab path to current working directory
CWD_PATH = os.getcwd()

# Path to frozen detection graph .pb file, which contains the model that is used
# for object detection.
PATH_TO_CKPT = os.path.join(CWD_PATH,MODEL_NAME,'frozen_inference_graph.pb')

# Path to label map file
PATH_TO_LABELS = os.path.join(CWD_PATH,'training','label_map.pbtxt')

# Path to video
PATH_TO_VIDEO = os.path.join(CWD_PATH,VIDEO_NAME)

# Number of classes the object detector can identify
NUM_CLASSES = 2

# Load the label map.
# Label maps map indices to category names, so that when our convolution
# network predicts `5`, we know that this corresponds to `king`.
# Here we use internal utility functions, but anything that returns a
# dictionary mapping integers to appropriate string labels would be fine
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)

# Load the Tensorflow model into memory.
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)

# Define input and output tensors (i.e. data) for the object detection classifier

# Input tensor is the image
image_tensor = detection_graph.get_tensor_by_name('image_tensor:0')

# Output tensors are the detection boxes, scores, and classes
# Each box represents a part of the image where a particular object was detected
detection_boxes = detection_graph.get_tensor_by_name('detection_boxes:0')

# Each score represents level of confidence for each of the objects.
# The score is shown on the result image, together with the class label.
detection_scores = detection_graph.get_tensor_by_name('detection_scores:0')
detection_classes = detection_graph.get_tensor_by_name('detection_classes:0')

# Number of objects detected
num_detections = detection_graph.get_tensor_by_name('num_detections:0')

# Open video file
video = cv2.VideoCapture('C:/Mohammad/tensorflow1/models/research/object_detection/test.mp4')
frame_width = int(video.get(3))
frame_height = int(video.get(4))
out = cv2.VideoWriter(FILE_OUTPUT, cv2.VideoWriter_fourcc('M', 'J', 'P', 'G'),
                      10, (frame_width, frame_height))


frame_index = 0
while(video.isOpened()):


    # 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_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})

    # Draw the results of the detection (aka 'visulaize the results')
    vis_util.visualize_boxes_and_labels_on_image_array(
        frame,
        frame_index,
        np.squeeze(boxes),
        np.squeeze(classes).astype(np.int32),
        np.squeeze(scores),
        category_index,
        use_normalized_coordinates=True,
        line_thickness=8,
        min_score_thresh=0.80)
    #frame_index +=1
    if ret == True:
        out.write(frame)

# writing coordinates
    coordinates = vis_util.return_coordinates(
        frame,
        frame_index,
        np.squeeze(boxes),
        np.squeeze(classes).astype(np.int32),
        np.squeeze(scores),
        category_index,
        use_normalized_coordinates=True,
        line_thickness=8,
        min_score_thresh=0.80)

    for coordinate in coordinates:
        (ymin, ymax, xmin, xmax, acc, classification) = coordinate
        height = ymax - ymin
        width = xmax -xmin
        crop = frame[ymin:ymin+height, xmin:xmin+width]
        path = 'C:/Mohammad/A_Mohammad laptop/Tensorflow From C Drive/tensorflow1/models/research/object_detection/savecroped'
        cv2.imwrite(os.path.join(path,'crop%d.jpg' %frame_index), frame)

        textfile = open('filename_string' + ".json", "a")
        textfile.write(json.dumps(coordinates))
        textfile.write("\n")
    frame_index =frame_index+1

# txt file
    #textfile = open('filename_string' + ".txt", "w")

    #textfile.write(str(coordinates))
    #textfile.write("\n")

   # All the results have been drawn on the frame, so it's time to display it.
    cv2.imshow('Object detector', frame)


    # Press 'q' to quit
    if cv2.waitKey(1) == ord('q'):
        break

# Clean up
video.release()
cv2.destroyAllWindows()

Answer 1

所以最好的方法是定义一个空列表并存储所有检测结果（边界框的坐标），并使用 centroid 类的更新函数来跟踪对象。也不要忘记在每次检测后清空列表，否则你会得到一个对象的这么多 ID！

Answer 2

import os
DATA_DIR = os.path.join(os.getcwd(), 'data')
MODELS_DIR = os.path.join(DATA_DIR, 'models')
for dir in [DATA_DIR, MODELS_DIR]:
    if not os.path.exists(dir):
        os.mkdir(dir)


import tarfile
import threading
import urllib.request
import six
from playsound import playsound


def soundPlay():
    playsound('Sound.mp3')


def get_Center(x, y, w, h): 
    x1 = int(w / 2)
    y1 = int(h / 2)
    cx = x + x1
    cy = y + y1
    return cx, cy


ROI = 50

offset = 12 



# Download and extract model
MODEL_DATE = '20200711'
MODEL_NAME = 'ssd_mobilenet_v1_fpn_640x640_coco17_tpu-8'
MODEL_TAR_FILENAME = MODEL_NAME + '.tar.gz'
MODELS_DOWNLOAD_BASE = 'http://download.tensorflow.org/models/object_detection/tf2/'
MODEL_DOWNLOAD_LINK = MODELS_DOWNLOAD_BASE + MODEL_DATE + '/' + MODEL_TAR_FILENAME
PATH_TO_MODEL_TAR = os.path.join(MODELS_DIR, MODEL_TAR_FILENAME)
PATH_TO_CKPT = os.path.join(MODELS_DIR, os.path.join(MODEL_NAME, 'checkpoint/'))
PATH_TO_CFG = os.path.join(MODELS_DIR, os.path.join(MODEL_NAME, 'pipeline.config'))


# Download labels file
LABEL_FILENAME = 'mscoco_label_map.pbtxt'
LABELS_DOWNLOAD_BASE = \
    'https://raw.githubusercontent.com/tensorflow/models/master/research/object_detection/data/'
PATH_TO_LABELS = os.path.join(MODELS_DIR, os.path.join(MODEL_NAME, LABEL_FILENAME))


# %%
# Load the model
# ~~~~~~~~~~~~~~
# Next we load the downloaded model

os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'    # Suppress TensorFlow logging
import tensorflow as tf
from object_detection.utils import label_map_util
from object_detection.utils import config_util
from utils import visualization_utils as viz_utils
from object_detection.builders import model_builder

tf.get_logger().setLevel('ERROR')           # Suppress TensorFlow logging (2)

# Enable GPU dynamic memory allocation
gpus = tf.config.experimental.list_physical_devices('GPU')
for gpu in gpus:
    tf.config.experimental.set_memory_growth(gpu, True)

# Load pipeline config and build a detection model
configs = config_util.get_configs_from_pipeline_file(PATH_TO_CFG)
model_config = configs['model']
detection_model = model_builder.build(model_config=model_config, is_training=False)

# Restore checkpoint
ckpt = tf.compat.v2.train.Checkpoint(model=detection_model)
ckpt.restore(os.path.join(PATH_TO_CKPT, 'ckpt-0')).expect_partial()

@tf.function
def detect_fn(image):
    """Detect objects in image."""

    image, shapes = detection_model.preprocess(image)
    prediction_dict = detection_model.predict(image, shapes)
    detections = detection_model.postprocess(prediction_dict, shapes)

    return detections, prediction_dict, tf.reshape(shapes, [-1])


# %%
# Load label map data (for plotting)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Label maps correspond index numbers to category names, so that when our convolution network
# predicts `5`, we know that this corresponds to `airplane`.  Here we use internal utility
# functions, but anything that returns a dictionary mapping integers to appropriate string labels
# would be fine.
category_index = label_map_util.create_category_index_from_labelmap(PATH_TO_LABELS,
                                                                    use_display_name=True)

# %%
# Define the video stream
# ~~~~~~~~~~~~~~~~~~~~~~~
# We will use `OpenCV <https://pypi.org/project/opencv-python/>`_ to capture the video stream
# generated by our webcam. For more information you can refer to the `OpenCV-Python Tutorials <https://opencv-python-tutroals.readthedocs.io/en/latest/py_tutorials/py_gui/py_video_display/py_video_display.html#capture-video-from-camera>`_
import cv2

cap = cv2.VideoCapture("video_road.mp4")

import numpy as np

while True:
    # Read frame from camera
    ret, image_np = cap.read()

    # Expand dimensions since the model expects images to have shape: [1, None, None, 3]
    image_np_expanded = np.expand_dims(image_np, axis=0)



    input_tensor = tf.convert_to_tensor(np.expand_dims(image_np, 0), dtype=tf.float32)
    detections, predictions_dict, shapes = detect_fn(input_tensor)

    label_id_offset = 1
    image_np_with_detections = image_np.copy()

    cv2.line(image_np_with_detections, (50 , ROI), (1200 , ROI), (255, 0, 0), 3)  # Line



    detected_objects = [category_index.get(value) for index,value in enumerate((detections['detection_classes'][0].numpy() + label_id_offset).astype(int)) if detections['detection_scores'][0, index].numpy() > 0.50]


    if len(detected_objects) > 0:

        for obj in detected_objects[0]:

            if detected_objects[0]['name'] == "car":

                viz_utils.visualize_boxes_and_labels_on_image_array(
                    image_np_with_detections,
                    detections['detection_boxes'][0].numpy(),
                    (detections['detection_classes'][0].numpy() + label_id_offset).astype(int),
                    detections['detection_scores'][0].numpy(),
                    category_index,
                    use_normalized_coordinates=True,
                    max_boxes_to_draw=200,
                    min_score_thresh=.50,
                    agnostic_mode=False)
                
                boxes = np.squeeze(detections['detection_boxes'][0].numpy())
                scores = np.squeeze(detections['detection_scores'][0].numpy())

                min_score_thresh = 0.50

                bboxes = boxes[scores > min_score_thresh]
            
                im_width, im_height = image_np.shape[1::-1]

                for box in bboxes:
                    ymin, xmin, ymax, xmax = box

                    ymin = ymin * im_height
                    xmin = xmin * im_width
                    ymax = ymax * im_height
                    xmax = xmax * im_width

                    x = xmin
                    y = ymin
                    w = xmax - xmin
                    h = ymax - ymin

                    mid_point = get_Center(int(x), int(y), int(w),int(h))
                    cv2.circle(image_np_with_detections, (mid_point[0], mid_point[1]), 3, (0, 0, 255), -1)
                    cv2.rectangle(image_np_with_detections, (int(x),int(y)), (int(x) + int(w), int(y) + int(h)), (0,255,0), 2)
                    cv2.putText(image_np_with_detections, 'Car {}'.format(round(100*scores[0])), (int(x) + int(w), int(y) + int(h)), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 1, cv2.LINE_AA)
                    

                    if mid_point[1] < (ROI + offset) and mid_point[1] > (ROI - offset):
                        thread1 = threading.Thread(target = soundPlay)
                        thread1.start()

                        cv2.line(image_np_with_detections, (50 , ROI), (1200 , ROI), (0, 0, 255), 3)
                        cv2.imwrite("Cache_Image.jpg", image_np_with_detections)

    # Display output
    cv2.imshow('object detection', cv2.resize(image_np_with_detections, (800, 600)))

    if cv2.waitKey(25) & 0xFF == ord('q'):
        break

cap.release()
cv2.destroyAllWindows()