capradeepgujaran
commited on
Commit
•
c3b34ed
1
Parent(s):
92928c5
Update app.py
Browse files
app.py
CHANGED
@@ -47,46 +47,6 @@ class SafetyMonitor:
|
|
47 |
img_base64 = base64.b64encode(buffered.getvalue()).decode('utf-8')
|
48 |
return f"data:image/jpeg;base64,{img_base64}"
|
49 |
|
50 |
-
def get_scene_context(self, image):
|
51 |
-
"""Analyze the scene context."""
|
52 |
-
try:
|
53 |
-
image_url = self.encode_image(image)
|
54 |
-
completion = self.client.chat.completions.create(
|
55 |
-
model=self.model_name,
|
56 |
-
messages=[
|
57 |
-
{
|
58 |
-
"role": "user",
|
59 |
-
"content": [
|
60 |
-
{
|
61 |
-
"type": "text",
|
62 |
-
"text": """Analyze this workplace image and identify key areas and elements. Include:
|
63 |
-
1. Worker locations and activities
|
64 |
-
2. Equipment and machinery
|
65 |
-
3. Materials and storage
|
66 |
-
4. Access routes and paths
|
67 |
-
5. Hazardous areas
|
68 |
-
|
69 |
-
Format each observation as:
|
70 |
-
- Element: specific location in image"""
|
71 |
-
},
|
72 |
-
{
|
73 |
-
"type": "image_url",
|
74 |
-
"image_url": {
|
75 |
-
"url": image_url
|
76 |
-
}
|
77 |
-
}
|
78 |
-
]
|
79 |
-
}
|
80 |
-
],
|
81 |
-
temperature=0.3,
|
82 |
-
max_tokens=200,
|
83 |
-
stream=False
|
84 |
-
)
|
85 |
-
return completion.choices[0].message.content
|
86 |
-
except Exception as e:
|
87 |
-
print(f"Scene analysis error: {str(e)}")
|
88 |
-
return ""
|
89 |
-
|
90 |
def analyze_frame(self, frame):
|
91 |
"""Perform safety analysis on the frame."""
|
92 |
if frame is None:
|
@@ -113,14 +73,6 @@ class SafetyMonitor:
|
|
113 |
}
|
114 |
}
|
115 |
]
|
116 |
-
},
|
117 |
-
{
|
118 |
-
"role": "assistant",
|
119 |
-
"content": "I'll analyze the image for safety hazards and provide the location and description of each concern."
|
120 |
-
},
|
121 |
-
{
|
122 |
-
"role": "user",
|
123 |
-
"content": "Please format each observation as: - <location>position:detailed safety concern</location>"
|
124 |
}
|
125 |
],
|
126 |
temperature=0.7,
|
@@ -136,12 +88,6 @@ class SafetyMonitor:
|
|
136 |
"""Convert textual position to coordinates."""
|
137 |
height, width = image_shape[:2]
|
138 |
|
139 |
-
# Parse position for spatial information
|
140 |
-
position = position.lower()
|
141 |
-
|
142 |
-
# Base coordinates (full image)
|
143 |
-
x1, y1, x2, y2 = 0, 0, width, height
|
144 |
-
|
145 |
# Define regions
|
146 |
regions = {
|
147 |
'center': (width//3, height//3, 2*width//3, 2*height//3),
|
@@ -158,18 +104,9 @@ class SafetyMonitor:
|
|
158 |
'middle': (0, height//3, width, 2*height//3)
|
159 |
}
|
160 |
|
161 |
-
#
|
162 |
-
|
163 |
-
|
164 |
-
for region, coords in regions.items():
|
165 |
-
if region in position:
|
166 |
-
words = region.split('-')
|
167 |
-
matches = sum(1 for word in words if word in position)
|
168 |
-
if matches > max_match:
|
169 |
-
max_match = matches
|
170 |
-
best_match = coords
|
171 |
-
|
172 |
-
return best_match if best_match else (x1, y1, x2, y2)
|
173 |
|
174 |
def draw_observations(self, image, observations):
|
175 |
"""Draw bounding boxes and labels for safety observations."""
|
@@ -184,6 +121,7 @@ class SafetyMonitor:
|
|
184 |
|
185 |
# Get coordinates for this observation
|
186 |
x1, y1, x2, y2 = self.get_region_coordinates(obs['location'], image.shape)
|
|
|
187 |
|
188 |
# Draw rectangle
|
189 |
cv2.rectangle(image, (x1, y1), (x2, y2), color, 2)
|
|
|
47 |
img_base64 = base64.b64encode(buffered.getvalue()).decode('utf-8')
|
48 |
return f"data:image/jpeg;base64,{img_base64}"
|
49 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
50 |
def analyze_frame(self, frame):
|
51 |
"""Perform safety analysis on the frame."""
|
52 |
if frame is None:
|
|
|
73 |
}
|
74 |
}
|
75 |
]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
76 |
}
|
77 |
],
|
78 |
temperature=0.7,
|
|
|
88 |
"""Convert textual position to coordinates."""
|
89 |
height, width = image_shape[:2]
|
90 |
|
|
|
|
|
|
|
|
|
|
|
|
|
91 |
# Define regions
|
92 |
regions = {
|
93 |
'center': (width//3, height//3, 2*width//3, 2*height//3),
|
|
|
104 |
'middle': (0, height//3, width, 2*height//3)
|
105 |
}
|
106 |
|
107 |
+
# Ensure the region name from the model output matches one of our predefined regions
|
108 |
+
position = position.lower()
|
109 |
+
return regions.get(position, (0, 0, width, height)) # Default to full image if no match
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
110 |
|
111 |
def draw_observations(self, image, observations):
|
112 |
"""Draw bounding boxes and labels for safety observations."""
|
|
|
121 |
|
122 |
# Get coordinates for this observation
|
123 |
x1, y1, x2, y2 = self.get_region_coordinates(obs['location'], image.shape)
|
124 |
+
print(f"Drawing box at coordinates: ({x1}, {y1}, {x2}, {y2}) for {obs['description']}")
|
125 |
|
126 |
# Draw rectangle
|
127 |
cv2.rectangle(image, (x1, y1), (x2, y2), color, 2)
|