def check_for_tasks():
# Face model training commands
if database.bool_run_train_face_model():
print('[INFO] Training face model')
# Extract embeddings
extract_embeddings.extract_embeddings(cwd_path=os.getcwd(),
input_confidence=0.5)
# Train model
train_model.train_model(cwd_path=os.getcwd())
# Detection tasks for re processing
detection_tasks = database.get_detection_tasks()
for row in detection_tasks:
# Get db row fields
id = row[0]
label = row[1]
file_name_cropped = row[2]
print('[INFO] Processing re-recognition task ' + file_name_cropped)
# Path
output_image = output_root_folder_path + label + '/' + file_name_cropped
# Label based detection if not detected earlier
detection_result, color = detection_utils.detect(
label=label,
crop_image_file_path_name_extension=output_image,
file_name=label + '_' + file_name_cropped,
output_file_name=None)
# Write new detection result
database.update_detection_task_result(id, detection_result)
def app():
# Do work
sr_work_records = database.get_super_resolution_images_to_compute()
sr_image_objects = []
for row in sr_work_records:
# Get db row fields
id = row[0]
label = row[1]
cropped_file_name = row[2]
detection_result = row[3]
# Construct paths
input_image = output_root_folder_path + label + '/' + cropped_file_name
output_image_path = output_root_folder_path + label + '/' + 'super_resolution/'
output_image = output_image_path + cropped_file_name
# Check path existence
Path(output_image_path).mkdir(parents=True, exist_ok=True)
# Make objects
sr_image_object = SrFile.SrFile(id, label, cropped_file_name,
input_image, output_image,
detection_result, '')
sr_image_objects.append(sr_image_object)
# Super resolution image
if len(sr_image_objects) > 0:
# Process super resolution images
sr_image_objects = infer_oi.process_super_resolution_images(
sr_image_objects=sr_image_objects,
max_width=max_width,
max_height=max_height)
# Process results
for sr_image_object in sr_image_objects:
# Label based detection if not detected earlier
if is_null(sr_image_object.detection_result):
sr_image_object.detection_result, sr_image_object.color = detection_utils.detect(
label=sr_image_object.label,
crop_image_file_path_name_extension=sr_image_object.
output_image,
file_name=sr_image_object.image_name,
output_file_name=sr_image_object.label + '_' +
sr_image_object.image_name,
use_rotation=True)
# Try to detect color
try:
sr_image_object.color = vehicle_color_detect.detect_color(
sr_image_object.output_image)
except Exception as e:
print(e)
# Write database, row no longer processed later
database.update_super_resolution_row_result(
sr_image_object.detection_result, sr_image_object.color,
sr_image_object.image_name, sr_image_object.id)
else:
print('No new sr image objects to process')
def analyze_image(image_object, bool_move_processed):
try:
# Load Yolo
net = cv2.dnn.readNet(models_path + "yolov4.weights",
models_path + "yolov4.cfg")
classes = []
with open(models_path + "coco.names", "r") as f:
classes = [line.strip() for line in f.readlines()]
layer_names = net.getLayerNames()
output_layers = [
layer_names[i[0] - 1] for i in net.getUnconnectedOutLayers()
]
colors = np.random.uniform(0, 255, size=(len(classes), 3))
# Loading image
img = cv2.imread(image_object.file_path + image_object.file_name)
# Storing original img and dimensions
original_h, original_w, original_c = img.shape
img_full_size = img.copy()
img = cv2.resize(img, None, fx=0.4, fy=0.4)
img_box = img.copy()
height, width, channels = img.shape
# Detecting objects
blob = cv2.dnn.blobFromImage(img,
0.00392, (608, 608), (0, 0, 0),
True,
crop=False)
net.setInput(blob)
outs = net.forward(output_layers)
# Showing information's on the screen
class_ids = []
confidences = []
boxes = []
original_img_boxes = []
for out in outs:
for detection in out:
scores = detection[5:]
class_id = np.argmax(scores)
confidence = scores[class_id]
if confidence > 0.5:
# Object detected
center_x = int(detection[0] * width)
center_y = int(detection[1] * height)
w = int(detection[2] * width)
h = int(detection[3] * height)
# Rectangle coordinates
x = int(center_x - w / 2)
y = int(center_y - h / 2)
boxes.append([x, y, w, h])
confidences.append(float(confidence))
class_ids.append(class_id)
# For original image boxes
center_x_ = int(detection[0] * original_w)
center_y_ = int(detection[1] * original_h)
w_ = int(detection[2] * original_w)
h_ = int(detection[3] * original_h)
# Rectangle coordinates
x_ = int(center_x_ - w_ / 2)
y_ = int(center_y_ - h_ / 2)
original_img_boxes.append([x_, y_, w_, h_])
# When we perform the detection, it happens that we have more boxes for the same object, so we should use
# another function to remove this “noise”. It’s called Non maximum suppression.
indexes = cv2.dnn.NMSBoxes(boxes, confidences, 0.5, 0.4)
# Output file name
out_file_name = image_object.year + '_' + image_object.month + '_' + image_object.day + '_' + image_object.hours + '_' + image_object.minutes + '_' + image_object.seconds
# We finally extract all the information's and show them on the screen.
font = cv2.FONT_HERSHEY_PLAIN
for i in range(len(boxes)):
if i in indexes:
detection_result = ''
color = ''
x, y, w, h = filter_negative_values(boxes[i])
x_, y_, w_, h_ = filter_negative_values(original_img_boxes[i])
# roi = img_box[y:y + h, x:x + w]
roi_full_image = img_full_size[y_:y_ + h_, x_:x_ + w_]
label = str(classes[class_ids[i]])
color = colors[i]
cv2.rectangle(img, (x, y), (x + w, y + h), color, 1)
cv2.putText(img, label, (x, y + 20), font, 2, color, 2)
if bool_label_is_ignored(label) is False:
# Small image naming and output path
crop_image_name_extension = out_file_name + '_' + str(
i) + image_object.file_extension
crop_image_file_path_name_extension = output_path + label + '/' + crop_image_name_extension
# Write small image
try:
Path(output_path + label + '/').mkdir(parents=True,
exist_ok=True)
cv2.imwrite(crop_image_file_path_name_extension,
roi_full_image)
except Exception as e:
print(e)
# Label based detection
detection_result, color = detection_utils.detect(
label=label,
crop_image_file_path_name_extension=
crop_image_file_path_name_extension,
file_name=out_file_name,
output_file_name=label + '_' + out_file_name + '_' +
str(i) + image_object.file_extension,
use_rotation=True)
# Insert database intelligence
try:
database.insert_value(
image_object.name, label, image_object.file_path,
image_object.file_name, image_object.year,
image_object.month, image_object.day,
image_object.hours, image_object.minutes,
image_object.seconds, crop_image_name_extension,
detection_result, color)
except Exception as e:
print(e)
# Move processed image
if bool_move_processed:
shutil.move(
image_object.file_path + image_object.file_name,
image_object.file_path + 'processed/' + image_object.file_name)
# Show preview
if cv2_imshow_enabled is True:
cv2.imshow("Image", img)
cv2.waitKey(1) # no freeze, refreshes for a millisecond
# Write full detection image
# only write if there's boxes available
if len(boxes) > 0:
try:
Path(object_detection_image_path).mkdir(parents=True,
exist_ok=True)
cv2.imwrite(
object_detection_image_path + '/' +
image_object.file_name + image_object.file_extension, img)
except Exception as e:
print(e)
# cv2.waitKey(0)
# cv2.destroyAllWindows()
except AssertionError as e:
print(e)