-
Notifications
You must be signed in to change notification settings - Fork 2
/
reach_obj_det_YoloV4.py
366 lines (314 loc) · 13.4 KB
/
reach_obj_det_YoloV4.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
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
import time
import rospy
import importlib
baxter=importlib.import_module("baxter-python3.baxter")
face=importlib.import_module("baxter-python3.faces")
import cv2
import numpy as np
import math
import argparse
from baxter_core_msgs.msg import EndpointState
PI = 3.141592
WIDTH = 960
HEIGHT = 600
DISPLAY_FACE=True
unreachable_count=0
garabbed=False
model_list = ["yolov4","yolov4-new","yolov4x-mish","yolov4-p6","yolov4-colors"]
parser = argparse.ArgumentParser()
parser.add_argument('-m', '--model', type=str, default='yolov4-colors', help='Model desired among '+str(model_list))
parser.add_argument('-o', '--object', type=str, default='apple', help='Object to reach and pick')
parser.add_argument('-a', '--arm', type=str, default='left', help='Arm, left or right')
args = parser.parse_args()
OBJECT_DESIRED = args.object
side = args.arm
print("[INFO] loading model...")
if args.model == "yolov4":
#Load net
modelConfig = "./models/yolov4.cfg"
modelWeigths = "./models/yolov4.weights"
net = cv2.dnn.readNetFromDarknet(modelConfig, modelWeigths)
print("Net Loaded: {}".format(args.model))
with open('./models/coco.names', 'r') as f:
classes = f.read().splitlines()
print("Classes: {}".format(len(classes)))
conf_threshold = 0.1
nms_threshold = 0.6 #lower=stronger
elif args.model == "yolov4-new":
#Load net
modelConfig = "./models/yolov4_new.cfg"
modelWeigths = "./models/yolov4_new.weights"
net = cv2.dnn.readNetFromDarknet(modelConfig, modelWeigths)
print("Net Loaded: {}".format(args.model))
with open('./models/coco.names', 'r') as f:
classes = f.read().splitlines()
print("Classes: {}".format(len(classes)))
#suggested
conf_threshold = 0.35
nms_threshold = 0.03 #lower=stronger
elif args.model == "yolov4x-mish":
#Load net
modelConfig = "./models/yolov4x-mish.cfg"
modelWeigths = "./models/yolov4x-mish.weights"
net = cv2.dnn.readNetFromDarknet(modelConfig, modelWeigths)
print("Net Loaded: {}".format(args.model))
with open('./models/coco.names', 'r') as f:
classes = f.read().splitlines()
print("Classes: {}".format(len(classes)))
#suggested
conf_threshold = 0.35
nms_threshold = 0.01 #lower=stronger
elif args.model == "yolov4-p6":
#Load net
modelConfig = "./models/yolov4-p6-1280x1280.cfg"
modelWeigths = "./models/yolov4-p6-1280x1280.weights"
net = cv2.dnn.readNetFromDarknet(modelConfig, modelWeigths)
print("Net Loaded: {}".format(args.model))
with open('./models/coco.names', 'r') as f:
classes = f.read().splitlines()
print("Classes: {}".format(len(classes)))
#suggested
conf_threshold = 0.35
nms_threshold = 0.01 #lower=stronger
elif args.model == "yolov4-colors":
#Load net
modelConfig = "./models/yolov4-colors.cfg"
modelWeigths = "./models/yolov4-colors.weights"
net = cv2.dnn.readNetFromDarknet(modelConfig, modelWeigths)
print("Net Loaded: {}".format(args.model))
with open('./models/colors.names', 'r') as f:
classes = f.read().splitlines()
print("Classes: {}".format(len(classes)))
#suggested
conf_threshold = 0.35
nms_threshold = 0.6 #lower=stronger
else:
print("[Error] Model passed not present, choose between: {}".format(model_list))
exit()
np.random.seed(42) #to generate the same colors
colors = np.random.randint(0, 255, size=(len(classes), 3), dtype='uint8')
print("Colors generated: "+str(colors.shape[0]))
# function to get the output layer names
# in the architecture
def get_output_layers(net):
layer_names = net.getLayerNames()
output_layers = []
for i in net.getUnconnectedOutLayers():
i = i[0] #DEPENDING ON YOUR OPENCV VERSION delete this line and similars in the code (i it's not an array anymore)
output_layers.append(layer_names[i-1])
return output_layers
# function to draw bounding box on the detected object with class name
def draw_bounding_box(img_yolo, class_id, confidence, x, y, x_plus_w, y_plus_h):
label = str(classes[class_id])
# Preparing colour for current bounding box
color = [int(j) for j in colors[class_id]]
cv2.rectangle(img_yolo, (x,y), (x_plus_w,y_plus_h), color, 2)
text_box_current = '{}: {:.2f}'.format(label, confidence)
if y<5:(x,y)=(x+15, y+30) #label position not out of the image
cv2.putText(img_yolo, text_box_current, (x-6,y-6), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0,0,0), 2)
cv2.putText(img_yolo, text_box_current, (x-5,y-5), cv2.FONT_HERSHEY_SIMPLEX, 0.7, color, 2)
print("[INFO] starting robot...")
np.random.seed()
rospy.init_node("testing"+str(np.random.randint(100))) #random node name so multiple can exist
rospy.sleep(2.0)
robot = baxter.BaxterRobot(rate=100, arm=side)
face._set_look(robot, side, DISPLAY_FACE)
rospy.sleep(2.0)
robot._set_camera(camera_name=side+"_hand_camera", state=True, width=WIDTH, height=HEIGHT, fps=30)
robot.set_robot_state(True)
rospy.sleep(2.0)
print("[INFO] calibrate gripper...")
robot.gripper_calibrate()
rospy.sleep(2.0)
robot.gripper_release()
#display face
face._set_look(robot, side+"_down", DISPLAY_FACE)
print("[INFO] moving in position...")
print(robot.move_to_neutral())
face._set_look(robot, side, DISPLAY_FACE)
print(robot.move_to_zero())
face._set_look(robot, "frontal", DISPLAY_FACE)
data = np.array(list(robot._cam_image.data), dtype=np.uint8)
middle_point = np.array([WIDTH/2, HEIGHT/2])
#move over the table
if side=="left":
pos_x = 0.8203694373186249
pos_y = 0.08642622598662506
else:
pos_x = 0.7456267492841516
pos_y = -0.18863639477015234
pos_z = 0.28462916699929078
ori_x = 0.011154239796145276
ori_y = 0.9989687054009745
ori_z = -0.006554586552752852
ori_w = 0.06499079561397379
face._set_look(robot, "down", DISPLAY_FACE)
robot.set_cartesian_position([pos_x, pos_y, pos_z], [ori_x, ori_y, ori_z, ori_w])
print("[INFO] getting image stream and passing to DNN...")
while not rospy.is_shutdown():
img = np.array(list(robot._cam_image.data), dtype=np.uint8)
img = img.reshape(int(HEIGHT), int(WIDTH), 4)
img = img[:, :, :3].copy()
#Passing image to DNN
#gray_img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# create input blob
blob = cv2.dnn.blobFromImage(img, 1/255, (640,640), (0,0,0), True, crop=False)
# set input blob for the network
net.setInput(blob)
# run inference through the network
# and gather predictions from output layers
start = time.time()
outs = net.forward(get_output_layers(net))
print('\nPrediction took {:.5f} seconds'.format(time.time() - start))
# initialization
class_ids = []
confidences = []
boxes = []
# for each detetion from each output layer
# get the confidence, class id, bounding box params
# and ignore weak detections (confidence < conf_threshold)
for out in outs:
for detection in out:
scores = detection[5:]
class_id = np.argmax(scores)
confidence = scores[class_id]
if confidence > conf_threshold:
center_x = int(detection[0] * WIDTH)
center_y = int(detection[1] * HEIGHT)
w = int(detection[2] * WIDTH)
h = int(detection[3] * HEIGHT)
x = center_x - w / 2
y = center_y - h / 2
class_ids.append(class_id)
confidences.append(float(confidence))
boxes.append([x, y, w, h])
#apply non-max suppression
indices = cv2.dnn.NMSBoxes(boxes, confidences, conf_threshold, nms_threshold)
print("Detections: "+str(indices.shape[0])) if len(indices)!=0 else print("No Detections")
#set gripper center, different from image center, should be in between the tips of the gripper, a little down.
gripper_delta_x=0
gripper_delta_y=0
if side=="left": #cameras in hands can be slightly diffenent angles, so center of gripper in the images is different
gripper_delta_x=40
gripper_delta_y=-60
elif side=="right":
gripper_delta_x=40
gripper_delta_y=-100
#center for the calculations
center_object_x = round(WIDTH/2)+gripper_delta_x # usually the gripper center is a little on the right of the image of camera
center_object_y = round(HEIGHT/2)+gripper_delta_y # usually the gripper is a litte up compared to the camera
#object reset: not detected and in the center
object_present=False
objects_list=[]
object_x=0
object_y=0
# go through the detections remaining
# after nms and draw bounding box
for i in indices:
i = i[0] #DEPENDING ON YOUR OPENCV VERSION delete this line and similars in the code (i it's not an array anymore)
box = boxes[i]
x = box[0]
y = box[1]
w = box[2]
h = box[3]
draw_bounding_box(img, class_ids[i], confidences[i], round(x), round(y), round(x+w), round(y+h))
#save (X Y) object if present
if classes[class_ids[i]] == OBJECT_DESIRED:
object_present=True
object_x = round(x+(w/2))
object_y = round(y+(h/2))
objects_list.append((object_x,object_y))
print("{} found at: {} {}, size: {} {}".format(classes[class_ids[i]],object_x,object_y, w,h))
cv2.putText(img, "X", (object_x,object_y), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0,0,255), 2)
cv2.putText(img, "O", (center_object_x,center_object_y), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0,0,255), 3)
#object is selected as the closest one if there are many
if len(objects_list) > 1:
print("Many found, finding the closest")
closest_distance = float('inf')
for x, y in objects_list:
# Calculate Euclidean distance between the current point and the target point
distance = math.sqrt((x - center_object_x) ** 2 + (y - center_object_y) ** 2)
if distance < closest_distance:
closest_distance = distance
object_x, object_y = (x, y)
#put different color mark on object selected
cv2.putText(img, "X", (object_x,object_y), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0,255,0), 2)
#check infrared distance
if robot._ir_range.range > robot._ir_range.min_range and robot._ir_range.range < robot._ir_range.max_range:
current_range = robot._ir_range.range
distance_str= "Dist: {:0.2f}".format(robot._ir_range.range)
print(distance_str)
cv2.putText(img, distance_str, (50,50), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0,0,255), 2)
else:
current_range = 9999
print("Range sensor out of limits")
cv2.putText(img, "Dist: OUT", (50,50), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0,0,255), 2)
#display image
robot._set_display_data(cv2.resize(img, (1024,600)))
#if too close, grab
if current_range < 0.16 and not garabbed:
print("[info] Gripper CLOSE enough and object present, GRABBING without more movements")
face._set_face(robot, "determined", DISPLAY_FACE)
#grab
garabbed = True
robot.gripper_grip()
rospy.sleep(2.0)
#move
robot.set_cartesian_position([pos_x, pos_y, pos_z], [ori_x, ori_y, ori_z, ori_w])
robot.move_to_zero()
rospy.sleep(1.0)
robot.gripper_release()
garabbed=False
rospy.sleep(2.0)
robot.set_cartesian_position([pos_x, pos_y, pos_z], [ori_x, ori_y, ori_z, ori_w])
#if present and not close enough: move towards it
if object_present and not garabbed:
#get current arm position
msg = rospy.wait_for_message("/robot/limb/"+side+"/endpoint_state", EndpointState)
p = msg.pose.position
q = msg.pose.orientation
#compute deviation in image
delta_x_pixel=center_object_x - object_x
delta_y_pixel=center_object_y - object_y
print("DELTA PIXELS: {} and {}".format(delta_x_pixel, delta_y_pixel))
#compute movement robot
delta_x=0
delta_y=0
delta_z=0
delta_movement=0.05
if current_range < 0.25: #if close to something move less
delta_movement = 0.02
#if it's too on the side in X direction in the image move the robot on Y
if delta_x_pixel>40:
delta_y = delta_movement
elif delta_x_pixel<-40:
delta_y = -delta_movement
#if it's too on the side in Y direction in the image move the robot on X
if delta_y_pixel>40:
delta_x = delta_movement
elif delta_y_pixel<-40:
delta_x = -delta_movement
#if no horizontal movement the obj is centered, move down
if delta_y==0 and delta_x ==0:
delta_z = -delta_movement
#move
print("DELTA MOVEMENT X:{} Y:{} Z:{}".format(delta_x, delta_y, delta_z))
movement_valid = robot.set_cartesian_position([p.x+delta_x, p.y+delta_y, p.z+delta_z], [q.x, q.y, q.z, q.w], override_current_movement=False)
if movement_valid:
print("[info] Movement OK")
unreachable_count=0
elif not movement_valid and unreachable_count<4:
unreachable_count=unreachable_count+1
print("[info] Movement Unreachable count: {}".format(unreachable_count))
elif not movement_valid and unreachable_count>3:
face._set_face(robot, "worried", DISPLAY_FACE)
#set to origin
print("[info] Moving to Origin")
robot.set_cartesian_position([pos_x, pos_y, pos_z], [ori_x, ori_y, ori_z, ori_w])
#else:
#if enough time passed look around
#sleep
robot.rate.sleep()
#out of the cycle
print(robot.move_to_neutral())
robot.set_robot_state(False)