forked from CocaineCong/gesture-recognizion
-
Notifications
You must be signed in to change notification settings - Fork 0
/
main.py
294 lines (277 loc) · 13.7 KB
/
main.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
"""
输出带操作
"""
from models.yolov3 import Yolov3
from models.resnet import resnet50
from utils.utils import process_data, scale_coords, process_coordinate, \
draw_bd_handpose, get_k, plot_one_box, load_classes, parse_data_cfg
from utils.nms import non_max_suppression
import torch
import numpy as np
import cv2
import math
import os
from javaPredict import load_java_pkg, java_close
from winOs.win_cmd import *
from models.experimental import attempt_load
def get_hand_data(im0, mirror=True):
if mirror:
im0 = cv2.flip(im0, 1, dst=None) # 水平镜像处理
img = process_data(im0, 416) # 数据预处理（用于传入yolo网络识别手的位置）
# 图片检测
img = torch.from_numpy(img).unsqueeze(0).to(device)
# pred, _ = yolov5_model(img)
pred = yolo_model(img)[0]
# 非极大值抑制nms
detections = non_max_suppression(pred, conf_thres, nms_thres)[0]
if detections is None or len(detections) == 0:
return im0, 0 # 未检测到手，直接跳过手势判断
# 将结果映射到原图
detections[:, :4] = scale_coords(416, detections[:, :4], im0.shape).round()
results = []
for *coordinate, conf, cls_conf, cls in detections:
x1, x2, y1, y2 = process_coordinate(coordinate, im0.shape)
hand_img = im0[y1:y2, x1:x2]
img_width = hand_img.shape[1]
img_height = hand_img.shape[0]
# 输入图片预处理
hand_img = cv2.resize(hand_img, (256, 256), interpolation=cv2.INTER_CUBIC)
hand_img = hand_img.astype(np.float32)
hand_img = (hand_img - 128.0) / 256.0
hand_img = hand_img.transpose(2, 0, 1)
hand_img = torch.from_numpy(hand_img)
hand_img = hand_img.unsqueeze_(0)
if use_cuda:
hand_img = hand_img.cuda() # (bs, 3, h, w)
pre_ = resnet50_model(hand_img.float()) # 模型推理
output = pre_.cpu().detach().numpy()
output = np.squeeze(output)
pts_hand = {} # 构建关键点连线可视化结构
point = []
for i in range(int(output.shape[0] / 2)):
x = (output[i * 2 + 0] * float(img_width)) + x1
y = (output[i * 2 + 1] * float(img_height)) + y1
point.append((x, y))
# cv2.putText(im0, str(i), (int(x), int(y)), cv2.FONT_HERSHEY_SIMPLEX, 0.8, (255, 255, 0), 2)
# 绘制关键点
cv2.circle(im0, (int(x), int(y)), 3, (255, 50, 60), -1)
cv2.circle(im0, (int(x), int(y)), 1, (255, 150, 180), -1)
# pts_hand[str(i)] = {}
pts_hand[str(i)] = {
"x": x,
"y": y,
}
draw_bd_handpose(im0, pts_hand, 0, 0) # 绘制关键点连线
flag = 0
hand_label = "hand"
if abs(180 - abs(get_k(point[0], point[3]) - get_k(point[3], point[4])) * 180 / math.pi) > 135:
hand_label += "-1"
flag += 1
if abs(180 - abs(get_k(point[0], point[6]) - get_k(point[6], point[8])) * 180 / math.pi) > 135:
hand_label += "-2"
flag += 10
if abs(180 - abs(get_k(point[0], point[10]) - get_k(point[10], point[12])) * 180 / math.pi) > 135:
hand_label += "-3"
flag += 100
if abs(180 - abs(get_k(point[0], point[14]) - get_k(point[14], point[16])) * 180 / math.pi) > 135:
hand_label += "-4"
flag += 1000
if abs(180 - abs(get_k(point[0], point[18]) - get_k(point[18], point[20])) * 180 / math.pi) > 135:
hand_label += "-5"
flag += 10000
if flag == 0 or flag == 1: # 握拳动作因人而异
hand_label = "0"
elif flag == 10 or flag == 11:
hand_label = "1"
else:
hand_label = "2"
label = '%s %.2f' % (hand_label, conf)
plot_one_box(coordinate, im0, label=label, color=(100, 0, 100), line_thickness=2)
results.append({
"x": x1,
"y": y1,
"w": x2 - x1,
"h": y2 - y1,
"label": hand_label
})
return im0, results
if __name__ == '__main__':
from config import conf_thres, nms_thres, cmd, data_cfg, \
resnet50_model_path, yolov3_model_path, yolo_choose, \
yolov5s_model_path, yolov5sm_model_path
use_cuda = torch.cuda.is_available()
device = torch.device("cuda:0" if use_cuda else "cpu")
# 加载resnet50模型
resnet50_model = resnet50(num_classes=42, img_size=256)
resnet50_model = resnet50_model.to(device)
resnet50_model.eval()
if os.access(resnet50_model_path, os.F_OK):
chkpt = torch.load(resnet50_model_path, map_location=device)
resnet50_model.load_state_dict(chkpt)
else:
raise Exception("resnet50模型权重文件丢失，无法继续进行。")
# 加载yolov5s模型
if yolo_choose == "yolov3":
# 加载yolov3模型
classes = load_classes(parse_data_cfg(data_cfg)['names'])
num_classes = len(classes)
anchors = [(10, 13), (16, 30), (33, 23), (30, 61), (62, 45), (59, 119), (116, 90), (156, 198),
(373, 326)] # yolov5先验框
yolo_model = Yolov3(num_classes=num_classes, anchors=anchors)
yolo_model = yolo_model.to(device)
yolo_model.eval()
if os.access(yolov3_model_path, os.F_OK):
yolo_model.load_state_dict(torch.load(yolov3_model_path, map_location=device)['model'])
else:
raise Exception("yolov3模型权重文件丢失，无法继续进行。")
elif yolo_choose == "yolov5sm":
if os.access(yolov5sm_model_path, os.F_OK):
yolo_model = attempt_load(yolov5sm_model_path, map_location=device)
else:
raise Exception("yolov5sm模型权重文件丢失，无法继续进行。")
else:
if os.access(yolov5s_model_path, os.F_OK):
yolo_model = attempt_load(yolov5s_model_path, map_location=device)
else:
raise Exception("yolov5sm模型权重文件丢失，无法继续进行。")
# 加载java环境,并返回预测对象
recognizer, recorderInterface = load_java_pkg()
# video_capture = cv2.VideoCapture(1, cv2.CAP_DSHOW)
video_capture = cv2.VideoCapture(0) # 选择默认摄像头
# video_capture = cv2.VideoCapture('./video/3.mp4') # 选择视频
isSet = 0
maxHeight = 0
setFrame = 15
middleLine = 240
hands = {}
d = {}
with torch.no_grad(): # 设置无梯度运行
while True:
ret, im0 = video_capture.read() # 读取图片流
if ret:
im0, result = get_hand_data(im0, mirror=True)
if result:
if isSet < setFrame and len(result) == 2:
"""初始化（需要setFrame帧内手势最大位置变化不大）"""
newMaxHeight = max(result[0]["y"], result[1]["y"])
if abs(maxHeight - newMaxHeight) > 20:
maxHeight = max(result[0]["y"], result[1]["y"])
isSet = 0
else:
isSet += 1
if isSet == setFrame:
maxHeight -= 40 # 略大保证终止
recognizer.setDockLevel(maxHeight)
middleLine = (result[0]["x"] + result[0]["w"] // 2 + result[1]["x"] + result[1]["w"] // 2) // 2 # 取两手中点为中线
hands = {}
print("初始化成功,maxHeight:", maxHeight, "middleLine:", middleLine)
else:
"""初始化完成后的判断"""
if not hands:
if len(result) > 1:
"""有多个手，取边缘两只手进行比较，根据相对位置判断左右手"""
if result[0]["x"] + result[0]["w"] // 2 < result[-1]["x"] + result[-1]["w"] // 2:
hands['0'] = result[0]
hands['1'] = result[1]
else:
hands['0'] = result[1]
hands['1'] = result[0]
else:
t1 = None
t2 = None
m1 = 999999
m2 = 999999
if hands.get("0", None): # 寻找与左手最近的手
v = hands.get("0")
for hand in result:
"""找出结果中和当前选择手最近的那个"""
l = (hand['x'] - v['x']) * (hand['x'] - v['x']) + (hand['y'] - v['y']) * (hand['y'] - v['y'])
if l < m1:
m1 = l
t1 = hand
if hands.get("1", None): # 寻找与右手最近的手
v = hands.get("1")
for hand in result:
"""找出结果中和当前选择手最近的那个"""
l = (hand['x'] - v['x']) * (hand['x'] - v['x']) + (hand['y'] - v['y']) * (hand['y'] - v['y'])
if l < m2:
m2 = l
t2 = hand
if t1 == t2:
if m1 < m2:
hands["0"] = t1
else:
hands["1"] = t2
else:
if t1:
hands["0"] = t1
if t2:
hands["1"] = t2
if hands:
for (k, v) in hands.items():
ans = recorderInterface.addOne(
v["x"],
v["y"],
v["w"],
v["h"],
int(v["label"]),
int(k)
)
if ans:
handpose = str(ans.getAction())
if handpose == "CLICK":
handpose = "点击"
elif handpose == "PAN":
if str(ans.getLocus()) == "LEFT":
handpose = "向左平移"
elif str(ans.getLocus()) == "RIGHT":
handpose = "向右平移"
elif handpose == "ZOOM":
if str(ans.getLocus()) == "IN":
handpose = "缩放"
elif str(ans.getLocus()) == "OUT":
handpose = "放大"
elif handpose == "GRAB":
handpose = "抓取"
elif handpose == "PUNCH":
handpose = "重置"
elif handpose == "ROTATE":
if str(ans.getLocus()) == "COUNTER_CLOCKWISE_ARC":
handpose = "逆时针旋转"
elif str(ans.getLocus()) == "CLOCKWISE_ARC":
handpose = "顺时针旋转"
if d.get("动作：" + handpose + ";手：" + str(ans.getHand())):
d["动作：" + handpose + ";手：" + str(ans.getHand())] += 1
else:
d["动作：" + handpose + ";手：" + str(ans.getHand())] = 1
print("动作：" + handpose + ";手：" + str(ans.getHand()))
if handpose == "重置":
isSet = 0
maxHeight = 0
hands = {}
recognizer.setDockLevel(maxHeight)
if cmd == 1:
cmdPPT(handpose)
elif cmd == 2:
cmdPic(handpose)
if isSet == setFrame:
"""画初始线"""
cv2.line(im0, (0, maxHeight), (im0.shape[1], maxHeight), (255, 0, 0), 3)
cv2.line(im0, (middleLine, 0), (middleLine, im0.shape[0]), (255, 0, 0), 3)
for (k, hand) in hands.items():
cv2.putText(im0, k, (hand["x"], hand['y']), cv2.FONT_HERSHEY_PLAIN, 2, (0, 255, 0), 3)
cv2.namedWindow('image', 0)
cv2.imshow("image", im0)
key = cv2.waitKey(1)
if key == 27:
break
elif key == 13:
isSet = 0
maxHeight = 0
hands = {}
recognizer.setDockLevel(maxHeight)
else:
break
cv2.destroyAllWindows()
java_close()
print(d)