def run(data_dir):
detector = Detector()
with open('data.tsv', 'r') as fp:
r = csv.reader(fp, delimiter='\t')
next(r)
for row in r:
photo_url = row[3]
basename = os.path.basename(photo_url)
path0 = f'{ord(basename[0]):02x}'
path1 = f'{ord(basename[1]):02x}'
path2 = f'{ord(basename[2]):02x}'
filepath = os.path.join(data_dir, 'images', path0, path1, path2,
basename)
print(f'processing {filepath} ...')
try:
result = detector.detect(cv2.imread(filepath))
if result is None:
print('detection failed.')
continue
outdir = os.path.join(data_dir, 'results', path0, path1, path2)
os.makedirs(outdir, exist_ok=True)
img = result['image']
del result['image']
result['meta'] = {
'face_id': row[0],
'photo_id': row[1],
'source_url': row[2],
'photo_url': row[3],
'posted_at': row[4],
'label_id': row[5],
'label_name': row[6],
}
name = os.path.splitext(basename)[0]
cv2.imwrite(os.path.join(outdir, f'{name}.png'), img)
with open(os.path.join(outdir, f'{name}.json'), 'w') as fp:
json.dump(result, fp, ensure_ascii=False)
except:
print(f'error!!: {filepath}')
time.sleep(1)
class CalcGridLayout(GridLayout):
filePath = StringProperty('')
weights = r'weights\best-100.pt'
grafic_list = []
label_list = []
weights_description = [
'Red entrenada 100 ciclos con dataset SVHN',
'Red entrenada 1000 ciclos con dataset SVHN',
'Red entrenada 1000 ciclos con dataset UNO Cards',
"Red entrenada 1000 ciclos con dataset SVHN con datos incrementados"
]
def __init__(self, **kwargs):
super(CalcGridLayout, self).__init__(**kwargs)
Window.bind(on_dropfile=self._on_file_drop)
self.detector = Detector(self.weights, self.filePath)
self.ids.model_label.text = self.weights_description[0]
def reduced_image(self):
pass
def _on_file_drop(self, window, file_path):
self.reset_canvas()
self.filePath = file_path.decode("utf-8")
self.ids.img.source = self.filePath
self.ids.img.reload()
self.ids.message_label.text = ' '
self.ids.start.visible = True
self.detector.change_source(self.filePath)
def change_weight(self, value):
self.ids.model_label.text = self.weights_description[value]
if value == 0:
self.weights = 'weights/best-100.pt'
if value == 1:
self.weights = 'weights/best-1000.pt'
if value == 2:
self.weights = 'weights/uno.pt'
if value == 3:
self.weights = 'weights/best-fused.pt'
self.detector.change_weights(self.weights)
def start_detection(self):
self.reset_canvas()
self.coordinates = self.detector.detect()
self.draw_rectangle(self.coordinates)
def reset_canvas(self):
for item in self.grafic_list:
self.canvas.remove(item)
for item in self.label_list:
self.ids.float_canvas.remove_widget(item)
for item in self.canvas.children:
if 'Line' in repr(item):
self.canvas.remove(item)
def draw_rectangle(self, coordinates):
print(coordinates)
full_size_x = self.ids.img.size[0]
full_size_y = self.ids.img.size[1]
center = (full_size_x / 2, full_size_y / 2)
image = Image.open(self.filePath)
image_width = int(image.size[0])
image_height = int(image.size[1])
for coordinate in coordinates:
rectangle_size_x = coordinate['bottom_right'][0] - coordinate[
'top_left'][0]
rectangle_size_y = coordinate['bottom_right'][1] - coordinate[
'top_left'][1]
relative_x_position = (center[0] - image_width / 2 +
coordinate['top_left'][0])
relative_y_position = (center[1] + image_height / 2 -
coordinate['top_left'][1] -
rectangle_size_y)
with self.canvas:
self.obj = InstructionGroup()
self.obj.add(Color(0, 1, 0, 0.3, mode="rgba"))
self.obj.add(
Line(rectangle=(relative_x_position, relative_y_position,
rectangle_size_x, rectangle_size_y),
width=3))
self.grafic_list.append(self.obj)
label_position_x = relative_x_position - center[
0] + rectangle_size_x / 2
label_position_y = relative_y_position - center[
1] + rectangle_size_y * 1.2
label = Label(text=coordinate['label'],
pos_hint={
'x': label_position_x / full_size_x,
'y': label_position_y / full_size_y
},
size=(self.ids.float_canvas.size[0],
self.ids.float_canvas.size[1]),
size_hint=(None, None),
bold=True,
outline_color=[1, 0, 0, 0.5],
outline_width=2)
self.ids.float_canvas.add_widget(label)
self.label_list.append(label)
videoframe = cv2.VideoCapture(videoname)
framenr = 0
if not videoframe.isOpened():
print("Error opening video stream or file")
while videoframe.isOpened():
ret, img = videoframe.read()
if not ret: # reached end of video
break
start = time.time()
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
detections = detector.detect(img)
det_list, crop_list = detector.get_detections(img, detections)
det_ids = reid.evaluate_query(crop_list)
detector.save_pic_with_detections(img, detections, det_ids, title=f"pic-{framenr:04}")
print(f"Elaborating frame {framenr} fps: {1 / (time.time() - start):.3}")
framenr += 1
print("Processing finished, saving video")
videoframe.release()
cv2.destroyAllWindows()
save_video(work_dir, output_dir, filename=opt.outname)
print(f"Video saved as {os.path.join(output_dir, opt.outname)}")
def worker():
provider = ServerAudioProvider(queue)
detector = Detector(provider)
detector.detect()
def annotate(cap,text_file_path):
global frameNum
frameNum = 0
cap.read()
_, first_frame = cap.read()
mask = toolMaskDrawer.main('',image=first_frame)
#need to massively change below lines
parameters = importlib.import_module("parameters.will1")
#parameters.red_range,parameters.white_range = automation.findColorRanges(first_frame,debug=True)
global detector
detector = Detector(mask,parameters)
global state
state = State(parameters)
#global current_positions
#current_positions = {i:(-1,-1) for i in range(10)}
global ball_infos
ball_infos = [BallInfo(state.getBall(i)) for i in range(10)]
global text_file
initial_frameNum = 0
if cont:
read_file = open(text_file_path,"r")
remember_line = ''
d = {}
while(True):
line = read_file.readline()
if line == '':
break
remember_line = line
_, d = read_annotation.processLine(line)
for k,v in d.items():
print k,v
ball_infos[k].ball.position = v
ball_infos[k].position = v
initial_frameNum = int(remember_line.split(':')[0])
text_file = open(text_file_path,"a")
else:
text_file = open(text_file_path,"w")
circles = detector.detect(first_frame)
if not cont:
for color in [Color.RED,Color.WHITE]:
colored_circles = sorted(circles[color], key=lambda c: c.x)#copied code from tracker v3
balls = state.red_balls if color == Color.RED else state.white_balls
for i,ball in enumerate(balls):
if i >= len(colored_circles):
break
ball.position = colored_circles[i].center()
while(initial_frameNum!=frameNum):
ret,frame = cap.read()
frameNum+=1
while(cap.isOpened()):
global frame, frameNum
ret, frame = cap.read()
frameNum += 1
redraw()
if not handleInput():
break
write()
cv2.destroyAllWindows()
def main():
yolo = Detector()
# Definition of the parameters
max_cosine_distance = 0.3
nn_budget = None
nms_max_overlap = 1.0
# deep_sort
model_filename = 'model_data/mars-small128.pb'
encoder = gdet.create_box_encoder(model_filename, batch_size=1)
metric = nn_matching.NearestNeighborDistanceMetric("cosine",
max_cosine_distance,
nn_budget)
tracker = Tracker(metric)
writeVideo_flag = True
video_capture = cv2.VideoCapture("town.avi")
if writeVideo_flag:
# Define the codec and create VideoWriter object
w = int(video_capture.get(3))
h = int(video_capture.get(4))
fourcc = cv2.VideoWriter_fourcc(*'MJPG')
out = cv2.VideoWriter('output.avi', fourcc, 15, (w, h))
list_file = open('detection.txt', 'w')
frame_index = -1
fps = 0.0
maxframe = 1
nframe = 0
while True:
ret, frame = video_capture.read() # frame shape 640*480*3
if ret != True:
break
t1 = time.time()
detections = []
# image = Image.fromarray(frame)
nframe += 1
if (nframe >= maxframe):
boxs, obj = yolo.detect(frame)
print(len(boxs))
# print("box_num",len(boxs))
features = encoder(frame, boxs)
# score to 1.0 here).
detections = [
Detection(bbox, 1.0, feature)
for bbox, feature in zip(boxs, features)
]
# Run non-maxima suppression.
boxes = np.array([d.tlwh for d in detections])
scores = np.array([d.confidence for d in detections])
indices = preprocessing.non_max_suppression(
boxes, nms_max_overlap, scores)
detections = [detections[i] for i in indices]
tracker.predict()
tracker.update(detections)
# Call the tracker
nframe = 0
for track in tracker.tracks:
if not track.is_confirmed() or track.time_since_update > 1:
continue
bbox = track.to_tlbr()
# # s=recognize(img)
# try:
# if(len(s)>len(track.track_lpn)):
# track.track_lpn=s
# except Exception as e:
# print(e)
cv2.rectangle(frame, (int(bbox[0]), int(bbox[1])),
(int(bbox[2]), int(bbox[3])), (255, 255, 255), 2)
cv2.putText(frame, str(track.track_id),
(int(bbox[0]), int(bbox[1])), 0, 5e-3 * 200,
(0, 255, 0), 2)
# for det in detections:
# bbox = det.to_tlbr()
# cv2.rectangle(frame,(int(bbox[0]), int(bbox[1])), (int(bbox[2]), int(bbox[3])),(255,0,0), 2)
cv2.imshow('', frame)
if writeVideo_flag:
# save a frame
out.write(frame)
frame_index = frame_index + 1
list_file.write(str(frame_index) + ' ')
fps = (fps + (1. / (time.time() - t1))) / 2
print("fps= %f" % (fps))
# Press Q to stop!
if cv2.waitKey(1) & 0xFF == ord('q'):
break
video_capture.release()
if writeVideo_flag:
out.release()
list_file.close()
cv2.destroyAllWindows()
#infFile = "/home/leo/qj/object_detection/data/socr/inference_frcnn/frozen_inference_graph.pb"
#mapFile = "/home/leo/qj/object_detection/data/socr/pascal_label_map.pbtxt"
infFile = "/home/leo/qj/object_detection/data/socr/inference_ssd/frozen_inference_graph.pb"
mapFile = "/home/leo/qj/object_detection/data/socr/pascal_label_map.pbtxt"
#infFile = "/home/leo/qj/object_detection/data/Index/inference_ssd/frozen_inference_graph.pb"
#mapFile = "/home/leo/qj/object_detection/data/Index/pascal_label_map.pbtxt"
imFiles = [
"/home/leo/qj/object_detection/data/origin/socr_20180410/JPEGImages/im0505.jpg",
"/home/leo/qj/object_detection/data/origin/socr_20180410/JPEGImages/im0506.jpg",
"/home/leo/qj/object_detection/data/origin/socr_20180410/JPEGImages/im0507.jpg",
"/home/leo/qj/object_detection/data/origin/socr_20180410/JPEGImages/im0508.jpg"
]
detector = Detector('socr', infFile, mapFile)
for imFile in imFiles:
img_np = np.asarray(Image.open(imFile))
img_np.setflags(write=1)
s = time.time()
(image_np, tag_boxes, tag_scores, tag_classes,
tag_num) = detector.detect(img_np, visualize=False)
e = time.time()
print 'detecting:', imFile
print 'detect time is:', (e - s)
print 'tag_num:', tag_num
for c in tag_classes:
if tag_scores[c] >= 0.8:
print 'tag:(%s,%f)', c, tag_scores[c]
