forked from smellslikeml/ActionAI
-
Notifications
You must be signed in to change notification settings - Fork 0
/
legday.py
executable file
·88 lines (68 loc) · 2.46 KB
/
legday.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
#!/usr/bin/env python3
import time
import cv2
import numpy as np
import tensorflow as tf
from collections import deque
from keras.models import load_model
interpreter = tf.contrib.lite.Interpreter(model_path="./models/model.tflite")
interpreter.allocate_tensors()
input_details = interpreter.get_input_details()
output_details = interpreter.get_output_details()
def run_inference(image):
t = time.time()
interpreter.set_tensor(input_details[0]['index'], image)
interpreter.invoke()
elapsed = time.time() - t
return interpreter.get_tensor(output_details[0]['index']), elapsed
cam = cv2.VideoCapture(0)
M = np.float32([[1, 0, 0],[0, 1, -15]])
y_dict = {0:'stand', 1:'squat', 2:'deadlift'}
#Load pickled model
loaded_model = load_model('./models/lstm.h5')
N = 5
q = deque(maxlen=N)
def ffwd_n_avg(q):
for idx, val in enumerate(q[-1]):
if not val:
q[-1][idx] = q[-2][idx]
return q
count_dict = {'squat': 0, 'deadlift': 0, 'stand': 0}
while True:
ret_val, image = cam.read()
image = cv2.resize(image,(192,192),3)
image = image.reshape((1,192,192,3))
image = image.astype(np.float32)
output_data, elapsed = run_inference(image)
feature_vec = np.zeros(28)
for kp in range(14):
blf = output_data[:,:,:,kp]
max_idx = np.argmax(blf)
coords = divmod(max_idx, 96)
feature_vec[2*kp:2*kp+2] = coords
q.append(feature_vec)
if len(q) == N:
ff = np.expand_dims(np.concatenate(list(ffwd_n_avg(q)), axis=0).reshape(28, N), axis=0)
move_pred = loaded_model.predict(ff)
move_pred = np.argmax(move_pred[0])
count_dict[y_dict[move_pred]] += 1
im = np.sum(output_data, axis=3).reshape((96, 96))[:,::-1] * 2
im = cv2.warpAffine(im, M, (96, 96))
im = cv2.resize(im, (1016, 1856), interpolation=cv2.INTER_CUBIC)
cv2.putText(im,
"S: %i D: %i St: %i" % (count_dict['squat'], count_dict['deadlift'], count_dict['stand']),
(100, 1500), cv2.FONT_HERSHEY_SIMPLEX, 3.0,
(255, 255, 255), 10)
try:
cv2.putText(im,
"LegDay",
(100, 100), cv2.FONT_HERSHEY_SIMPLEX, 3.0,
(255, 255, 255), 10)
except:
pass
cv2.namedWindow('window', cv2.WND_PROP_FULLSCREEN)
cv2.setWindowProperty('window', cv2.WND_PROP_FULLSCREEN, cv2.WINDOW_FULLSCREEN)
cv2.imshow('window', im)
if cv2.waitKey(1) == 27:
break
cv2.destroyAllWindows()