def get_boxes(opts, outs, source_size, num=5, threshold=0.1):
assert box_constructor is not None, '`get_boxes` requires `darkflow`.'
h, w = source_size
boxes = [[] for _ in xrange(opts['classes'])]
opts['num'] = num
opts['thresh'] = threshold
opts['out_size'] = list(outs[0].shape)
results = box_constructor(opts, outs[0].copy())
for b in results:
max_idx = np.argmax(b.probs)
max_score = b.probs[max_idx]
if max_score > threshold:
x1 = int((b.x - b.w / 2) * w)
y1 = int((b.y - b.h / 2) * h)
x2 = int((b.x + b.w / 2) * w)
y2 = int((b.y + b.h / 2) * h)
if x1 < 0:
x1 = 0
if x2 > w - 1:
x2 = w - 1
if y1 < 0:
y1 = 0
if y2 > h - 1:
y2 = h - 1
boxes[max_idx].append((x1, y1, x2, y2, max_score))
for i in xrange(opts['classes']):
boxes[i] = np.asarray(boxes[i], dtype=np.float32)
return boxes
def findboxes(self, net_out):
# meta
meta = self.meta
boxes = list()
boxes = box_constructor(meta, net_out)
return boxes
def yolo_boxes(opts, outs, source_size):
assert box_constructor is not None, '`get_boxes` requires `darkflow`.'
h, w = source_size
threshold = opts['thresh']
boxes = [[] for _ in xrange(opts['classes'])]
results = box_constructor(opts, outs[0].copy())
for b in results:
max_indx = np.argmax(b.probs)
max_prob = b.probs[max_indx]
if max_prob > threshold:
left = int((b.x - b.w / 2) * w)
right = int((b.x + b.w / 2) * w)
top = int((b.y - b.h / 2) * h)
bot = int((b.y + b.h / 2) * h)
if left < 0:
left = 0
if right > w - 1:
right = w - 1
if top < 0:
top = 0
if bot > h - 1:
bot = h - 1
boxes[max_indx].append((left, top, right, bot, max_prob))
for i in xrange(opts['classes']):
boxes[i] = np.asarray(boxes[i], dtype=np.float32)
return boxes
def getResults(requestData):
with open(os.path.join("./built_graph/yolov2.meta"), 'r') as fp:
meta = json.load(fp)
imgdata = base64.b64decode(requestData["data"])
data = Image.open(io.BytesIO(imgdata))
image_np = cv2.cvtColor(np.array(data), cv2.COLOR_BGR2RGB)
h, w, c = meta['inp_size']
imsz = cv2.resize(image_np , (w, h))
imsz = imsz / 255.
imsz = imsz[:,:,::-1]
image_np_expanded = np.expand_dims(imsz, 0)
[h, w] = image_np.shape[:2]
headers = {'Accept': 'application/octet-stream',
'content-type': 'application/json'}
start_time1 = time.time()
r = requests.post(url = URL , data = json.dumps(requestData), headers = headers)
elapsed_time1 = time.time() - start_time
print("Time to get the results : ", elapsed_time1)
data = r.json()
data1 = ast.literal_eval(data)
out1 = data1["detection"]
boxes = box_constructor(meta, np.array(out1, dtype=np.float32))
#print(boxes)
boxesInfo = list()
#resultsForJSON = []
#colors = meta['colors']
threshold = meta['thresh']
for b in boxes:
max_indx = np.argmax(b.probs)
max_prob = b.probs[max_indx]
label = meta['labels'][max_indx]
if max_prob > threshold:
left = int ((b.x - b.w/2.) * w)
right = int ((b.x + b.w/2.) * w)
top = int ((b.y - b.h/2.) * h)
bot = int ((b.y + b.h/2.) * h)
if left < 0 : left = 0
if right > w - 1: right = w - 1
if top < 0 : top = 0
if bot > h - 1: bot = h - 1
mess = '{}'.format(label)
boxResults = (left, right, top, bot, mess, max_indx, max_prob)
print(boxResults)
def detect_frame(self, img):
# print("[Yitao] in yolo.py:detect_frame() is called...!")
# return self.tfnet.return_predict(img)
# Yitao-TLS-Begin
h, w, _ = img.shape
im = resize_input(img)
this_inp = np.expand_dims(im, 0)
request = predict_pb2.PredictRequest()
request.model_spec.name = 'darkflow_yolo'
request.model_spec.signature_name = 'predict_images'
request.inputs['input'].CopyFrom(
tf.contrib.util.make_tensor_proto(this_inp,
dtype=np.float32,
shape=this_inp.shape))
result = self.stub.Predict(request, 10.0) # 10 secs timeout
tmp = result.outputs['output']
tt = tensor_util.MakeNdarray(tmp)[0]
# print(tt[:,0,0])
# print(tt[:,0,1])
meta = {
'labels': [
'person', 'bicycle', 'car', 'motorbike', 'aeroplane', 'bus',
'train', 'truck', 'boat', 'traffic light', 'fire hydrant',
'stop sign', 'parking meter', 'bench', 'bird', 'cat', 'dog',
'horse', 'sheep', 'cow', 'elephant', 'bear', 'zebra',
'giraffe', 'backpack', 'umbrella', 'handbag', 'tie',
'suitcase', 'frisbee', 'skis', 'snowboard', 'sports ball',
'kite', 'baseball bat', 'baseball glove', 'skateboard',
'surfboard', 'tennis racket', 'bottle', 'wine glass', 'cup',
'fork', 'knife', 'spoon', 'bowl', 'banana', 'apple',
'sandwich', 'orange', 'broccoli', 'carrot', 'hot dog', 'pizza',
'donut', 'cake', 'chair', 'sofa', 'pottedplant', 'bed',
'diningtable', 'toilet', 'tvmonitor', 'laptop', 'mouse',
'remote', 'keyboard', 'cell phone', 'microwave', 'oven',
'toaster', 'sink', 'refrigerator', 'book', 'clock', 'vase',
'scissors', 'teddy bear', 'hair drier', 'toothbrush'
],
u'jitter':
0.3,
u'anchors': [
0.57273, 0.677385, 1.87446, 2.06253, 3.33843, 5.47434, 7.88282,
3.52778, 9.77052, 9.16828
],
u'random':
1,
'colors': [(254, 254, 254), (254, 254, 127), (254, 254, 0),
(254, 254, -127), (254, 254, -254), (254, 127, 254),
(254, 127, 127), (254, 127, 0), (254, 127, -127),
(254, 127, -254), (254, 0, 254), (254, 0, 127),
(254, 0, 0), (254, 0, -127), (254, 0, -254),
(254, -127, 254), (254, -127, 127), (254, -127, 0),
(254, -127, -127), (254, -127, -254), (254, -254, 254),
(254, -254, 127), (254, -254, 0), (254, -254, -127),
(254, -254, -254), (127, 254, 254), (127, 254, 127),
(127, 254, 0), (127, 254, -127), (127, 254, -254),
(127, 127, 254), (127, 127, 127), (127, 127, 0),
(127, 127, -127), (127, 127, -254), (127, 0, 254),
(127, 0, 127), (127, 0, 0), (127, 0, -127),
(127, 0, -254), (127, -127, 254), (127, -127, 127),
(127, -127, 0), (127, -127, -127), (127, -127, -254),
(127, -254, 254), (127, -254, 127), (127, -254, 0),
(127, -254, -127), (127, -254, -254), (0, 254, 254),
(0, 254, 127), (0, 254, 0), (0, 254, -127),
(0, 254, -254), (0, 127, 254),
(0, 127, 127), (0, 127, 0), (0, 127, -127),
(0, 127, -254), (0, 0, 254), (0, 0, 127), (0, 0, 0),
(0, 0, -127), (0, 0, -254), (0, -127, 254),
(0, -127, 127), (0, -127, 0), (0, -127, -127),
(0, -127, -254), (0, -254, 254), (0, -254, 127),
(0, -254, 0), (0, -254, -127), (0, -254, -254),
(-127, 254, 254), (-127, 254, 127), (-127, 254, 0),
(-127, 254, -127), (-127, 254, -254)],
u'num':
5,
u'thresh':
self.threshold,
'inp_size': [608, 608, 3],
u'bias_match':
1,
'out_size': [19, 19, 425],
'model':
'/home/yitao/Documents/fun-project/darknet-repo/darkflow/cfg/yolo.cfg',
u'absolute':
1,
'name':
'yolo',
u'coord_scale':
1,
u'rescore':
1,
u'class_scale':
1,
u'noobject_scale':
1,
u'object_scale':
5,
u'classes':
80,
u'coords':
4,
u'softmax':
1,
'net': {
u'hue': 0.1,
u'saturation': 1.5,
u'angle': 0,
u'decay': 0.0005,
u'learning_rate': 0.001,
u'scales': u'.1,.1',
u'batch': 1,
u'height': 608,
u'channels': 3,
u'width': 608,
u'subdivisions': 1,
u'burn_in': 1000,
u'policy': u'steps',
u'max_batches': 500200,
u'steps': u'400000,450000',
'type': u'[net]',
u'momentum': 0.9,
u'exposure': 1.5
},
'type':
u'[region]'
}
boxes = list()
boxes = box_constructor(meta, tt)
# for box in boxes:
# print("(%s, %s, %s, %s) with class_num = %s, probs = %s" % (str(box.x), str(box.y), str(box.w), str(box.h), str(box.class_num), str(box.probs)))
boxesInfo = list()
for box in boxes:
tmpBox = process_box(box, h, w, self.threshold, meta)
if tmpBox is None:
continue
boxesInfo.append({
"label": tmpBox[4],
"confidence": tmpBox[6],
"topleft": {
"x": tmpBox[0],
"y": tmpBox[2]
},
"bottomright": {
"x": tmpBox[1],
"y": tmpBox[3]
}
})
return boxesInfo
def findboxes_meta(meta, net_out):
boxes = list()
boxes = box_constructor(meta, net_out)
return boxes
