-
Notifications
You must be signed in to change notification settings - Fork 6
/
detection.py
187 lines (171 loc) · 8.67 KB
/
detection.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
import sys
from pathlib import Path, PurePath
sys.path.append("./models/research/object_detection/")
sys.path.append("./models/research/")
import os
import cv2
import numpy as np
import tensorflow as tf
from utils import label_map_util
from utils import visualization_utils as vis_util
from image_to_video_converter import images_to_video
from PIL import Image
class detector:
def __init__(self, model_directory):
model_path = os.path.join(model_directory, 'frozen_inference_graph.pb')
labelmap_path = os.path.join(model_directory, 'labelmap.pbtxt')
self.num_classes = 5
self.label_map = label_map_util.load_labelmap(labelmap_path)
self.categories = label_map_util.convert_label_map_to_categories(self.label_map,
max_num_classes=self.num_classes,
use_display_name=True)
self.category_index = label_map_util.create_category_index(self.categories)
self.detection_graph = tf.Graph()
with self.detection_graph.as_default():
self.od_graph_def = tf.GraphDef()
with tf.gfile.GFile(model_path, 'rb') as fid:
self.serialized_graph = fid.read()
self.od_graph_def.ParseFromString(self.serialized_graph)
tf.import_graph_def(self.od_graph_def, name='')
self.sess = tf.Session(graph=self.detection_graph)
# Define input and output tensors (i.e. data) for the object detection classifier
self.image_tensor = self.detection_graph.get_tensor_by_name('image_tensor:0')
self.detection_boxes = self.detection_graph.get_tensor_by_name('detection_boxes:0')
self.detection_scores = self.detection_graph.get_tensor_by_name('detection_scores:0')
self.detection_classes = self.detection_graph.get_tensor_by_name('detection_classes:0')
self.num_detections = self.detection_graph.get_tensor_by_name('num_detections:0')
def draw_boxes_for_image(self, frame, min_score_threshold):
frame_expanded = np.expand_dims(frame, axis=0)
(boxes, scores, classes, num) = self.sess.run(
[self.detection_boxes, self.detection_scores, self.detection_classes, self.num_detections],
feed_dict={self.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),
self.category_index,
use_normalized_coordinates=True,
line_thickness=2,
min_score_thresh=min_score_threshold)
"""
print("Self cateogry index")
print(self.category_index)
print("Score/Classes")
for box, score, cls in zip(np.squeeze(boxes), np.squeeze(scores),np.squeeze(classes).astype(np.int32)):
print(score, cls, self.category_index[cls])
"""
good_boxes = [box
for box, score, cls in zip(np.squeeze(boxes), np.squeeze(scores), np.squeeze(classes).astype(np.int32))
if score >= min_score_threshold and 'traffic' not in self.category_index[cls]['name']]
return frame, good_boxes
@staticmethod
def denormalize(box, width, height):
# Order taken from: https://www.tensorflow.org/api_docs/python/tf/image/draw_bounding_boxes
y_min, x_min, y_max, x_max = box[0], box[1], box[2], box[3]
x_min *= width
x_max *= width
y_min *= height
y_max *= height
return [x_min, x_max, y_min, y_max]
@staticmethod
def log_boxes(frame_number, boxes, ofile, width, height):
for box in boxes:
box = detector.denormalize(box, width, height)
# Cast float coordinates to integers
box = map(int, box)
box = [frame_number] + list(box)
line = "|".join(map(str, box))
print(line, file=ofile)
def process_image(self, video_name, frame_number, image_path,
min_score_threshold, output_path, save_images):
image = cv2.imread(image_path)
image_name = Path(image_path).stem
result_frame = None
# Set up logging file
log_name = os.path.join(output_path, f"{video_name}_log.txt")
with open(log_name, 'a') as log_file:
print("At Frame:", frame_number)
frame = np.array(image)
# Draw boxes
frame, boxes = self.draw_boxes_for_image(frame, min_score_threshold)
height, width, layers = frame.shape
# Log boxes
detector.log_boxes(frame_number, boxes, log_file, width, height)
# Save frame with boxes for output
result_frame = frame
# Save frame with boxes
if save_images:
frame_path = os.path.join(output_path, f"{video_name}_frame_{image_name}.png")
print("Saving image at", frame_path)
vis_util.save_image_array_as_png(frame, frame_path)
return result_frame
def process_image_folder(self, folder_path, min_score_threshold, output_path, save_images):
folder_name = Path(folder_path).stem
frames = []
file_names = os.listdir(folder_path)
file_names.sort()
for f in file_names:
image_path = os.path.join(folder_path, f)
if os.path.isfile(image_path):
frame_number = len(frames)
next_frame = self.process_image(folder_name, frame_number, image_path,
min_score_threshold, output_path, save_images)
frames.append(next_frame)
if save_images:
video_path = os.path.join(output_path, folder_name)
video = cv2.VideoCapture(video_path)
print("Saving video at", video_path)
images_to_video(frames, video_path, 30)
def process_video(self, video_path, min_score_threshold, output_path, save_images):
video_name = Path(video_path).stem
# Open video file
video = cv2.VideoCapture(video_path)
# Set up logging file
log_name = os.path.join(output_path, f"{video_name}_log.txt")
with open(log_name, 'a') as log_file:
frames = []
while(video.isOpened()):
ret, frame = video.read()
if not ret:
break
frame_number = len(frames)
print("At Frame:", frame_number)
# Draw boxes
frame, boxes = self.draw_boxes_for_image(frame, min_score_threshold)
height, width, layers = frame.shape
# Log boxes
detector.log_boxes(frame_number, boxes, log_file, width, height)
# Save frame with boxes
if save_images:
frame_path = os.path.join(output_path, f"{video_name}_frame_{frame_number}.png")
print("Saving image at", frame_path)
vis_util.save_image_array_as_png(frame, frame_path)
frames.append(frame)
# Save as video
if save_images:
out_video_path = os.path.join(output_path, f"{video_name}.avi")
print("Saving video at", out_video_path)
images_to_video(frames, out_video_path, 30)
# Clean up
video.release()
cv2.destroyAllWindows()
def default_detector():
det = detector("./trained_model/detectors/")
return det
def default_inference():
det = default_detector()
det.process_video("./data/SignaledJunctionRightTurn_1.avi", 0.70, "./output/temp/", False)
return det
if __name__ == "__main__":
import argparse
parser = argparse.ArgumentParser()
parser.add_argument('--model_path', help='Path to the frozen inference graph and labelmap files',
required=True)
parser.add_argument('--video_path', help='Path to the video', required=True)
parser.add_argument('--min_threshold', type=float, help='Minimum score threshold for a bounding box to be drawn', default=0.7)
parser.add_argument('--output_path', help='Path for storing output images and/or logs', required=True)
parser.add_argument('--save_images', action='store_true')
args = parser.parse_args()
det = detector(args.model_path)
det.process_video(args.video_path, args.min_threshold, args.output_path, args.save_images)