Update objec_detect_yolo.py
Browse files- objec_detect_yolo.py +170 -54
objec_detect_yolo.py
CHANGED
@@ -1,101 +1,217 @@
|
|
|
|
|
|
1 |
import cv2
|
2 |
import numpy as np
|
3 |
import os
|
4 |
from ultralytics import YOLO
|
5 |
import time
|
6 |
-
from typing import Tuple, Set
|
7 |
|
8 |
-
def
|
9 |
"""
|
10 |
-
Detects and tracks objects in a video using
|
11 |
|
12 |
Args:
|
13 |
-
path (str): Path to the input video file.
|
14 |
|
15 |
Returns:
|
16 |
-
Tuple[Set[str], str]:
|
17 |
- Set of unique detected object labels (e.g., {'Gun', 'Knife'})
|
18 |
- Path to the output annotated video with detection boxes and tracking IDs
|
|
|
|
|
|
|
|
|
19 |
"""
|
|
|
|
|
20 |
if not os.path.exists(path):
|
21 |
raise FileNotFoundError(f"Video file not found: {path}")
|
22 |
|
23 |
-
#
|
24 |
-
|
25 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
26 |
|
27 |
-
|
|
|
28 |
input_video_name = os.path.basename(path)
|
29 |
base_name = os.path.splitext(input_video_name)[0]
|
30 |
-
|
31 |
-
|
32 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
33 |
temp_output_path = os.path.join(output_dir, temp_output_name)
|
|
|
|
|
34 |
|
35 |
-
# Video
|
36 |
cap = cv2.VideoCapture(path)
|
37 |
if not cap.isOpened():
|
38 |
raise ValueError(f"Failed to open video file: {path}")
|
39 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
40 |
frame_width, frame_height = 640, 640
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
49 |
start = time.time()
|
50 |
-
|
|
|
51 |
|
52 |
while True:
|
53 |
ret, frame = cap.read()
|
54 |
-
if not ret:
|
55 |
break
|
56 |
|
57 |
-
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
|
67 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
68 |
out.write(annotated_frame)
|
69 |
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
-
|
74 |
-
|
75 |
-
else:
|
76 |
-
out.write(frame)
|
77 |
|
|
|
78 |
end = time.time()
|
|
|
|
|
79 |
cap.release()
|
80 |
out.release()
|
|
|
|
|
81 |
|
82 |
-
#
|
83 |
-
|
84 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
85 |
final_output_path = os.path.join(output_dir, final_output_name)
|
86 |
|
87 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
88 |
|
89 |
-
print(f"[INFO] Processing finished at {end:.2f} seconds")
|
90 |
-
print(f"[INFO] Total execution time: {end - start:.2f} seconds")
|
91 |
-
print(f"[INFO] Detected crimes: {detected_labels}")
|
92 |
-
print(f"[INFO] Annotated video saved at: {final_output_path}")
|
93 |
|
94 |
-
return
|
95 |
|
96 |
|
97 |
-
# Example usage (
|
98 |
# if __name__ == "__main__":
|
99 |
-
#
|
100 |
-
#
|
101 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
--- START OF FILE objec_detect_yolo.py ---
|
2 |
+
|
3 |
import cv2
|
4 |
import numpy as np
|
5 |
import os
|
6 |
from ultralytics import YOLO
|
7 |
import time
|
8 |
+
from typing import Tuple, Set, List
|
9 |
|
10 |
+
def detection(path: str) -> Tuple[Set[str], str]:
|
11 |
"""
|
12 |
+
Detects and tracks objects in a video using YOLOv8 model, saving an annotated output video.
|
13 |
|
14 |
Args:
|
15 |
+
path (str): Path to the input video file. Supports common video formats (mp4, avi, etc.)
|
16 |
|
17 |
Returns:
|
18 |
+
Tuple[Set[str], str]:
|
19 |
- Set of unique detected object labels (e.g., {'Gun', 'Knife'})
|
20 |
- Path to the output annotated video with detection boxes and tracking IDs
|
21 |
+
|
22 |
+
Raises:
|
23 |
+
FileNotFoundError: If input video doesn't exist
|
24 |
+
ValueError: If video cannot be opened/processed or output dir cannot be created
|
25 |
"""
|
26 |
+
|
27 |
+
# Validate input file exists
|
28 |
if not os.path.exists(path):
|
29 |
raise FileNotFoundError(f"Video file not found: {path}")
|
30 |
|
31 |
+
# --- Model Loading ---
|
32 |
+
# Construct path relative to this script file
|
33 |
+
model_path = os.path.join(os.path.dirname(__file__), "yolo", "best.pt")
|
34 |
+
if not os.path.exists(model_path):
|
35 |
+
raise FileNotFoundError(f"YOLO model file not found at: {model_path}")
|
36 |
+
try:
|
37 |
+
model = YOLO(model_path)
|
38 |
+
class_names = model.names # Get class label mappings
|
39 |
+
print(f"[INFO] YOLO model loaded from {model_path}. Class names: {class_names}")
|
40 |
+
except Exception as e:
|
41 |
+
raise ValueError(f"Failed to load YOLO model: {e}")
|
42 |
|
43 |
+
|
44 |
+
# --- Output Path Setup ---
|
45 |
input_video_name = os.path.basename(path)
|
46 |
base_name = os.path.splitext(input_video_name)[0]
|
47 |
+
# Sanitize basename to prevent issues with weird characters in filenames
|
48 |
+
safe_base_name = "".join(c if c.isalnum() or c in ('-', '_') else '_' for c in base_name)
|
49 |
+
|
50 |
+
# Define output directory relative to this script
|
51 |
+
# In HF Spaces, this will be inside the container's file system
|
52 |
+
output_dir = os.path.join(os.path.dirname(__file__), "results")
|
53 |
+
temp_output_name = f"{safe_base_name}_output_temp.mp4"
|
54 |
+
|
55 |
+
try:
|
56 |
+
os.makedirs(output_dir, exist_ok=True) # Create output dir if needed
|
57 |
+
if not os.path.isdir(output_dir):
|
58 |
+
raise ValueError(f"Path exists but is not a directory: {output_dir}")
|
59 |
+
except OSError as e:
|
60 |
+
raise ValueError(f"Failed to create or access output directory '{output_dir}': {e}")
|
61 |
+
|
62 |
temp_output_path = os.path.join(output_dir, temp_output_name)
|
63 |
+
print(f"[INFO] Temporary output will be saved to: {temp_output_path}")
|
64 |
+
|
65 |
|
66 |
+
# --- Video Processing Setup ---
|
67 |
cap = cv2.VideoCapture(path)
|
68 |
if not cap.isOpened():
|
69 |
raise ValueError(f"Failed to open video file: {path}")
|
70 |
|
71 |
+
# Get video properties for output writer
|
72 |
+
# Use source FPS if available and reasonable, otherwise default to 30
|
73 |
+
source_fps = cap.get(cv2.CAP_PROP_FPS)
|
74 |
+
output_fps = source_fps if 10 <= source_fps <= 60 else 30.0
|
75 |
+
|
76 |
+
# Process at a fixed resolution for consistency or use source resolution
|
77 |
+
# Using fixed 640x640 as potentially used during training/fine-tuning
|
78 |
frame_width, frame_height = 640, 640
|
79 |
+
# OR use source resolution (might require adjusting YOLO parameters if model expects specific size)
|
80 |
+
# frame_width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
|
81 |
+
# frame_height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
|
82 |
+
|
83 |
+
try:
|
84 |
+
out = cv2.VideoWriter(
|
85 |
+
temp_output_path,
|
86 |
+
cv2.VideoWriter_fourcc(*'mp4v'), # Use MP4 codec
|
87 |
+
output_fps,
|
88 |
+
(frame_width, frame_height)
|
89 |
+
)
|
90 |
+
if not out.isOpened():
|
91 |
+
# Attempt alternative codec if mp4v fails (less common)
|
92 |
+
print("[WARNING] mp4v codec failed, trying avc1...")
|
93 |
+
out = cv2.VideoWriter(
|
94 |
+
temp_output_path,
|
95 |
+
cv2.VideoWriter_fourcc(*'avc1'),
|
96 |
+
output_fps,
|
97 |
+
(frame_width, frame_height)
|
98 |
+
)
|
99 |
+
if not out.isOpened():
|
100 |
+
raise ValueError("Failed to initialize VideoWriter with mp4v or avc1 codec.")
|
101 |
+
|
102 |
+
except Exception as e:
|
103 |
+
cap.release() # Release capture device before raising
|
104 |
+
raise ValueError(f"Failed to create VideoWriter: {e}")
|
105 |
+
|
106 |
+
|
107 |
+
# --- Main Processing Loop ---
|
108 |
+
detected_classes: List[str] = [] # Track detected object class names
|
109 |
start = time.time()
|
110 |
+
frame_count = 0
|
111 |
+
print(f"[INFO] Video processing started...")
|
112 |
|
113 |
while True:
|
114 |
ret, frame = cap.read()
|
115 |
+
if not ret: # End of video or read error
|
116 |
break
|
117 |
|
118 |
+
frame_count += 1
|
119 |
+
# Resize frame BEFORE passing to model
|
120 |
+
resized_frame = cv2.resize(frame, (frame_width, frame_height))
|
121 |
+
|
122 |
+
try:
|
123 |
+
# Run YOLOv8 detection and tracking on the resized frame
|
124 |
+
results = model.track(
|
125 |
+
source=resized_frame, # Use resized frame
|
126 |
+
conf=0.7, # Confidence threshold
|
127 |
+
persist=True, # Maintain track IDs across frames
|
128 |
+
verbose=False # Suppress Ultralytics console output per frame
|
129 |
+
)
|
130 |
+
|
131 |
+
# Check if results are valid and contain boxes
|
132 |
+
if results and results[0] and results[0].boxes:
|
133 |
+
# Annotate the RESIZED frame with bounding boxes and track IDs
|
134 |
+
annotated_frame = results[0].plot() # plot() draws on the source image
|
135 |
+
|
136 |
+
# Record detected class names for this frame
|
137 |
+
for box in results[0].boxes:
|
138 |
+
if box.cls is not None: # Check if class ID is present
|
139 |
+
cls_id = int(box.cls[0]) # Get class index
|
140 |
+
if 0 <= cls_id < len(class_names):
|
141 |
+
detected_classes.append(class_names[cls_id])
|
142 |
+
else:
|
143 |
+
print(f"[WARNING] Detected unknown class ID: {cls_id}")
|
144 |
+
else:
|
145 |
+
# If no detections, use the original resized frame for the output video
|
146 |
+
annotated_frame = resized_frame
|
147 |
+
|
148 |
+
# Write the (potentially annotated) frame to the output video
|
149 |
out.write(annotated_frame)
|
150 |
|
151 |
+
except Exception as e:
|
152 |
+
print(f"[ERROR] Error processing frame {frame_count}: {e}")
|
153 |
+
# Write the unannotated frame to keep video timing consistent
|
154 |
+
out.write(resized_frame)
|
155 |
+
|
|
|
|
|
156 |
|
157 |
+
# --- Clean Up ---
|
158 |
end = time.time()
|
159 |
+
print(f"[INFO] Video processing finished. Processed {frame_count} frames.")
|
160 |
+
print(f"[INFO] Total processing time: {end - start:.2f} seconds")
|
161 |
cap.release()
|
162 |
out.release()
|
163 |
+
cv2.destroyAllWindows() # Close any OpenCV windows if they were opened
|
164 |
+
|
165 |
|
166 |
+
# --- Final Output Renaming ---
|
167 |
+
unique_detected_labels = set(detected_classes)
|
168 |
+
# Create a short string from labels for the filename
|
169 |
+
labels_str = "_".join(sorted(list(unique_detected_labels))).replace(" ", "_")
|
170 |
+
# Limit length to avoid overly long filenames
|
171 |
+
max_label_len = 50
|
172 |
+
if len(labels_str) > max_label_len:
|
173 |
+
labels_str = labels_str[:max_label_len] + "_etc"
|
174 |
+
if not labels_str: # Handle case where nothing was detected
|
175 |
+
labels_str = "no_detections"
|
176 |
+
|
177 |
+
final_output_name = f"{safe_base_name}_{labels_str}_output.mp4"
|
178 |
final_output_path = os.path.join(output_dir, final_output_name)
|
179 |
|
180 |
+
# Ensure final path doesn't already exist (rename might fail otherwise)
|
181 |
+
if os.path.exists(final_output_path):
|
182 |
+
os.remove(final_output_path)
|
183 |
+
|
184 |
+
try:
|
185 |
+
# Rename the temporary file to the final name
|
186 |
+
os.rename(temp_output_path, final_output_path)
|
187 |
+
print(f"[INFO] Detected object labels: {unique_detected_labels}")
|
188 |
+
print(f"[INFO] Annotated video saved successfully at: {final_output_path}")
|
189 |
+
except OSError as e:
|
190 |
+
print(f"[ERROR] Failed to rename {temp_output_path} to {final_output_path}: {e}")
|
191 |
+
# Fallback: return the temp path if rename fails but file exists
|
192 |
+
if os.path.exists(temp_output_path):
|
193 |
+
print(f"[WARNING] Returning path to temporary file: {temp_output_path}")
|
194 |
+
return unique_detected_labels, temp_output_path
|
195 |
+
else:
|
196 |
+
raise ValueError(f"Output video generation failed. No output file found.")
|
197 |
|
|
|
|
|
|
|
|
|
198 |
|
199 |
+
return unique_detected_labels, final_output_path
|
200 |
|
201 |
|
202 |
+
# # Example usage (commented out for library use)
|
203 |
# if __name__ == "__main__":
|
204 |
+
# test_video = input("Enter the local path to the video file: ").strip('"')
|
205 |
+
# if os.path.exists(test_video):
|
206 |
+
# try:
|
207 |
+
# print(f"[INFO] Processing video: {test_video}")
|
208 |
+
# labels, out_path = detection(test_video)
|
209 |
+
# print(f"\nDetection Complete.")
|
210 |
+
# print(f"Detected unique labels: {labels}")
|
211 |
+
# print(f"Output video saved to: {out_path}")
|
212 |
+
# except (FileNotFoundError, ValueError, Exception) as e:
|
213 |
+
# print(f"\nAn error occurred: {e}")
|
214 |
+
# else:
|
215 |
+
# print(f"Error: Input video file not found - {test_video}")
|
216 |
+
|
217 |
+
--- END OF FILE objec_detect_yolo.py ---
|