Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -36,74 +36,62 @@ def build_model(HIDDEN_UNITS=256, sequence_length=30, num_input_values=33*4, num
|
|
36 |
model.load_weights(load_dir)
|
37 |
return model
|
38 |
|
|
|
39 |
class VideoProcessor:
|
40 |
def __init__(self):
|
41 |
self.actions = np.array(['curl', 'press', 'squat'])
|
42 |
self.sequence_length = 30
|
43 |
self.colors = [(245,117,16), (117,245,16), (16,117,245)]
|
44 |
-
self.
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
#
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
55 |
|
56 |
-
@st.cache()
|
57 |
def draw_landmarks(self, image, results):
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
"""
|
62 |
-
self.mp_drawing.draw_landmarks(image, results.pose_landmarks, self.mp_pose.POSE_CONNECTIONS,
|
63 |
-
self.mp_drawing.DrawingSpec(color=(245,117,66), thickness=2, circle_radius=2),
|
64 |
-
self.mp_drawing.DrawingSpec(color=(245,66,230), thickness=2, circle_radius=2)
|
65 |
-
)
|
66 |
return image
|
67 |
-
|
68 |
@st.cache()
|
69 |
def extract_keypoints(self, results):
|
70 |
-
"""
|
71 |
-
Processes and organizes the keypoints detected from the pose estimation model
|
72 |
-
to be used as inputs for the exercise decoder models
|
73 |
-
|
74 |
-
"""
|
75 |
pose = np.array([[res.x, res.y, res.z, res.visibility] for res in results.pose_landmarks.landmark]).flatten() if results.pose_landmarks else np.zeros(33*4)
|
76 |
return pose
|
77 |
|
78 |
@st.cache()
|
79 |
def calculate_angle(self, a, b, c):
|
80 |
-
"""
|
81 |
-
Computes 3D joint angle inferred by 3 keypoints and their relative positions to one another
|
82 |
-
|
83 |
-
"""
|
84 |
a = np.array(a) # First
|
85 |
b = np.array(b) # Mid
|
86 |
c = np.array(c) # End
|
87 |
-
|
88 |
radians = np.arctan2(c[1]-b[1], c[0]-b[0]) - np.arctan2(a[1]-b[1], a[0]-b[0])
|
89 |
angle = np.abs(radians*180.0/np.pi)
|
90 |
-
|
91 |
if angle > 180.0:
|
92 |
angle = 360-angle
|
93 |
-
|
94 |
return angle
|
95 |
|
96 |
@st.cache()
|
97 |
def get_coordinates(self, landmarks, side, joint):
|
98 |
-
"""
|
99 |
-
Retrieves x and y coordinates of a particular keypoint from the pose estimation model
|
100 |
-
|
101 |
-
Args:
|
102 |
-
landmarks: processed keypoints from the pose estimation model
|
103 |
-
side: 'left' or 'right'. Denotes the side of the body of the landmark of interest.
|
104 |
-
joint: 'shoulder', 'elbow', 'wrist', 'hip', 'knee', or 'ankle'. Denotes which body joint is associated with the landmark of interest.
|
105 |
-
|
106 |
-
"""
|
107 |
coord = getattr(self.mp_pose.PoseLandmark, side.upper() + "_" + joint.upper())
|
108 |
x_coord_val = landmarks[coord.value].x
|
109 |
y_coord_val = landmarks[coord.value].y
|
@@ -111,100 +99,12 @@ class VideoProcessor:
|
|
111 |
|
112 |
@st.cache()
|
113 |
def viz_joint_angle(self, image, angle, joint):
|
114 |
-
"""
|
115 |
-
Displays the joint angle value near the joint within the image frame
|
116 |
-
|
117 |
-
"""
|
118 |
cv2.putText(image, str(int(angle)),
|
119 |
tuple(np.multiply(joint, [640, 480]).astype(int)),
|
120 |
cv2.FONT_HERSHEY_SIMPLEX, 0.5, (255, 255, 255), 2, cv2.LINE_AA
|
121 |
)
|
122 |
return
|
123 |
|
124 |
-
@st.cache()
|
125 |
-
def process(self, image):
|
126 |
-
"""
|
127 |
-
Function to process the video frame from the user's webcam and run the fitness trainer AI
|
128 |
-
|
129 |
-
Args:
|
130 |
-
image (numpy array): input image from the webcam
|
131 |
-
|
132 |
-
Returns:
|
133 |
-
numpy array: processed image with keypoint detection and fitness activity classification visualized
|
134 |
-
"""
|
135 |
-
# Pose detection model
|
136 |
-
image.flags.writeable = False
|
137 |
-
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
|
138 |
-
results = pose.process(image)
|
139 |
-
|
140 |
-
# Draw the hand annotations on the image.
|
141 |
-
image.flags.writeable = True
|
142 |
-
image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
|
143 |
-
self.draw_landmarks(image, results)
|
144 |
-
|
145 |
-
# Prediction logic
|
146 |
-
keypoints = self.extract_keypoints(results)
|
147 |
-
self.sequence.append(keypoints.astype('float32',casting='same_kind'))
|
148 |
-
self.sequence = self.sequence[-self.sequence_length:]
|
149 |
-
|
150 |
-
if len(self.sequence) == self.sequence_length:
|
151 |
-
res = model.predict(np.expand_dims(self.sequence, axis=0), verbose=0)[0]
|
152 |
-
# interpreter.set_tensor(self.input_details[0]['index'], np.expand_dims(self.sequence, axis=0))
|
153 |
-
# interpreter.invoke()
|
154 |
-
# res = interpreter.get_tensor(self.output_details[0]['index'])
|
155 |
-
|
156 |
-
self.current_action = self.actions[np.argmax(res)]
|
157 |
-
confidence = np.max(res)
|
158 |
-
|
159 |
-
# Erase current action variable if no probability is above threshold
|
160 |
-
if confidence < self.threshold:
|
161 |
-
self.current_action = ''
|
162 |
-
|
163 |
-
# Viz probabilities
|
164 |
-
image = self.prob_viz(res, image)
|
165 |
-
|
166 |
-
# Count reps
|
167 |
-
try:
|
168 |
-
landmarks = results.pose_landmarks.landmark
|
169 |
-
self.count_reps(
|
170 |
-
image, landmarks, mp_pose)
|
171 |
-
except:
|
172 |
-
pass
|
173 |
-
|
174 |
-
# Display graphical information
|
175 |
-
cv2.rectangle(image, (0,0), (640, 40), self.colors[np.argmax(res)], -1)
|
176 |
-
cv2.putText(image, 'curl ' + str(self.curl_counter), (3,30),
|
177 |
-
cv2.FONT_HERSHEY_SIMPLEX, 1, (255, 255, 255), 2, cv2.LINE_AA)
|
178 |
-
cv2.putText(image, 'press ' + str(self.press_counter), (240,30),
|
179 |
-
cv2.FONT_HERSHEY_SIMPLEX, 1, (255, 255, 255), 2, cv2.LINE_AA)
|
180 |
-
cv2.putText(image, 'squat ' + str(self.squat_counter), (490,30),
|
181 |
-
cv2.FONT_HERSHEY_SIMPLEX, 1, (255, 255, 255), 2, cv2.LINE_AA)
|
182 |
-
|
183 |
-
# return cv2.flip(image, 1)
|
184 |
-
return image
|
185 |
-
|
186 |
-
def process_video(self, video_file):
|
187 |
-
# Get the filename from the file object
|
188 |
-
filename = video_file.name
|
189 |
-
# Create a temporary file to write the contents of the uploaded video file
|
190 |
-
temp_file = open(filename, 'wb')
|
191 |
-
temp_file.write(video_file.read())
|
192 |
-
temp_file.close()
|
193 |
-
# Now we can open the video file using cv2.VideoCapture()
|
194 |
-
cap = cv2.VideoCapture(filename)
|
195 |
-
out_frames = []
|
196 |
-
while cap.isOpened():
|
197 |
-
ret, frame = cap.read()
|
198 |
-
if not ret:
|
199 |
-
break
|
200 |
-
frame_processed = self.process(frame)
|
201 |
-
out_frames.append(frame_processed)
|
202 |
-
cap.release()
|
203 |
-
# Remove the temporary file
|
204 |
-
os.remove(filename)
|
205 |
-
return out_frames
|
206 |
-
|
207 |
-
|
208 |
# Define Streamlit app
|
209 |
def main():
|
210 |
st.title("Real-time Exercise Detection")
|
|
|
36 |
model.load_weights(load_dir)
|
37 |
return model
|
38 |
|
39 |
+
# Define the VideoProcessor class for real-time video processing
|
40 |
class VideoProcessor:
|
41 |
def __init__(self):
|
42 |
self.actions = np.array(['curl', 'press', 'squat'])
|
43 |
self.sequence_length = 30
|
44 |
self.colors = [(245,117,16), (117,245,16), (16,117,245)]
|
45 |
+
self.pose = mp_pose.Pose(min_detection_confidence=0.5, min_tracking_confidence=0.5)
|
46 |
+
self.model = build_model()
|
47 |
+
|
48 |
+
def process_video(self, video_file):
|
49 |
+
# Get the filename from the file object
|
50 |
+
filename = video_file.name
|
51 |
+
# Create a temporary file to write the contents of the uploaded video file
|
52 |
+
temp_file = open(filename, 'wb')
|
53 |
+
temp_file.write(video_file.read())
|
54 |
+
temp_file.close()
|
55 |
+
# Now we can open the video file using cv2.VideoCapture()
|
56 |
+
cap = cv2.VideoCapture(filename)
|
57 |
+
out_frames = []
|
58 |
+
while cap.isOpened():
|
59 |
+
ret, frame = cap.read()
|
60 |
+
if not ret:
|
61 |
+
break
|
62 |
+
frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
|
63 |
+
results = self.pose.process(frame_rgb)
|
64 |
+
frame = self.draw_landmarks(frame, results)
|
65 |
+
out_frames.append(frame)
|
66 |
+
cap.release()
|
67 |
+
# Remove the temporary file
|
68 |
+
os.remove(filename)
|
69 |
+
return out_frames
|
70 |
|
|
|
71 |
def draw_landmarks(self, image, results):
|
72 |
+
mp_drawing.draw_landmarks(image, results.pose_landmarks, mp_pose.POSE_CONNECTIONS,
|
73 |
+
mp_drawing.DrawingSpec(color=(245,117,66), thickness=2, circle_radius=2),
|
74 |
+
mp_drawing.DrawingSpec(color=(245,66,230), thickness=2, circle_radius=2))
|
|
|
|
|
|
|
|
|
|
|
75 |
return image
|
76 |
+
|
77 |
@st.cache()
|
78 |
def extract_keypoints(self, results):
|
|
|
|
|
|
|
|
|
|
|
79 |
pose = np.array([[res.x, res.y, res.z, res.visibility] for res in results.pose_landmarks.landmark]).flatten() if results.pose_landmarks else np.zeros(33*4)
|
80 |
return pose
|
81 |
|
82 |
@st.cache()
|
83 |
def calculate_angle(self, a, b, c):
|
|
|
|
|
|
|
|
|
84 |
a = np.array(a) # First
|
85 |
b = np.array(b) # Mid
|
86 |
c = np.array(c) # End
|
|
|
87 |
radians = np.arctan2(c[1]-b[1], c[0]-b[0]) - np.arctan2(a[1]-b[1], a[0]-b[0])
|
88 |
angle = np.abs(radians*180.0/np.pi)
|
|
|
89 |
if angle > 180.0:
|
90 |
angle = 360-angle
|
|
|
91 |
return angle
|
92 |
|
93 |
@st.cache()
|
94 |
def get_coordinates(self, landmarks, side, joint):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
95 |
coord = getattr(self.mp_pose.PoseLandmark, side.upper() + "_" + joint.upper())
|
96 |
x_coord_val = landmarks[coord.value].x
|
97 |
y_coord_val = landmarks[coord.value].y
|
|
|
99 |
|
100 |
@st.cache()
|
101 |
def viz_joint_angle(self, image, angle, joint):
|
|
|
|
|
|
|
|
|
102 |
cv2.putText(image, str(int(angle)),
|
103 |
tuple(np.multiply(joint, [640, 480]).astype(int)),
|
104 |
cv2.FONT_HERSHEY_SIMPLEX, 0.5, (255, 255, 255), 2, cv2.LINE_AA
|
105 |
)
|
106 |
return
|
107 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
108 |
# Define Streamlit app
|
109 |
def main():
|
110 |
st.title("Real-time Exercise Detection")
|