Spaces:
Running
Running
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,113 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import cv2
|
3 |
+
import torch
|
4 |
+
from sahi import AutoDetectionModel
|
5 |
+
from sahi.predict import get_sliced_prediction
|
6 |
+
from motpy import Detection as MotpyDetection, MultiObjectTracker
|
7 |
+
import tempfile
|
8 |
+
|
9 |
+
# COCO class names (YOLOv8 default)
|
10 |
+
COCO_CLASSES = [
|
11 |
+
'person', 'bicycle', 'car', 'motorcycle', 'airplane', 'bus', 'train', 'truck', 'boat', 'traffic light',
|
12 |
+
'fire hydrant', 'stop sign', 'parking meter', 'bench', 'bird', 'cat', 'dog', 'horse', 'sheep', 'cow',
|
13 |
+
'elephant', 'bear', 'zebra', 'giraffe', 'backpack', 'umbrella', 'handbag', 'tie', 'suitcase', 'frisbee',
|
14 |
+
'skis', 'snowboard', 'sports ball', 'kite', 'baseball bat', 'baseball glove', 'skateboard', 'surfboard',
|
15 |
+
'tennis racket', 'bottle', 'wine glass', 'cup', 'fork', 'knife', 'spoon', 'bowl', 'banana', 'apple',
|
16 |
+
'sandwich', 'orange', 'broccoli', 'carrot', 'hot dog', 'pizza', 'donut', 'cake', 'chair', 'couch',
|
17 |
+
'potted plant', 'bed', 'dining table', 'toilet', 'tv', 'laptop', 'mouse', 'remote', 'keyboard', 'cell phone',
|
18 |
+
'microwave', 'oven', 'toaster', 'sink', 'refrigerator', 'book', 'clock', 'vase', 'scissors', 'teddy bear',
|
19 |
+
'hair drier', 'toothbrush'
|
20 |
+
]
|
21 |
+
|
22 |
+
|
23 |
+
model_path = "./yolo11n.pt"
|
24 |
+
detection_model = AutoDetectionModel.from_pretrained(
|
25 |
+
model_type='yolov8',
|
26 |
+
model_path=model_path,
|
27 |
+
confidence_threshold=0.3,
|
28 |
+
device='cpu' # Force CPU usage
|
29 |
+
)
|
30 |
+
|
31 |
+
def track_objects(video_path):
|
32 |
+
# Setup video processing
|
33 |
+
cap = cv2.VideoCapture(video_path)
|
34 |
+
width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
|
35 |
+
height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
|
36 |
+
fps = cap.get(cv2.CAP_PROP_FPS)
|
37 |
+
|
38 |
+
output_file = tempfile.NamedTemporaryFile(suffix='.mp4', delete=False)
|
39 |
+
output_path = output_file.name
|
40 |
+
|
41 |
+
fourcc = cv2.VideoWriter_fourcc(*'mp4v')
|
42 |
+
out = cv2.VideoWriter(output_path, fourcc, fps, (width, height))
|
43 |
+
|
44 |
+
tracker = MultiObjectTracker(
|
45 |
+
dt=0.1,
|
46 |
+
model_spec={
|
47 |
+
'order_pos': 1, 'dim_pos': 2,
|
48 |
+
'order_size': 0, 'dim_size': 2,
|
49 |
+
'q_var_pos': 5000., 'r_var_pos': 0.1
|
50 |
+
}
|
51 |
+
)
|
52 |
+
|
53 |
+
frame_count = 0
|
54 |
+
while cap.isOpened():
|
55 |
+
ret, frame = cap.read()
|
56 |
+
if not ret:
|
57 |
+
break
|
58 |
+
|
59 |
+
rgb_frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
|
60 |
+
result = get_sliced_prediction(
|
61 |
+
rgb_frame,
|
62 |
+
detection_model,
|
63 |
+
slice_height=512,
|
64 |
+
slice_width=512,
|
65 |
+
overlap_height_ratio=0.2,
|
66 |
+
overlap_width_ratio=0.2
|
67 |
+
)
|
68 |
+
|
69 |
+
detections = [
|
70 |
+
MotpyDetection(
|
71 |
+
box=[obj.bbox.minx, obj.bbox.miny, obj.bbox.maxx, obj.bbox.maxy],
|
72 |
+
score=obj.score.value,
|
73 |
+
class_id=obj.category.id
|
74 |
+
)
|
75 |
+
for obj in result.object_prediction_list
|
76 |
+
]
|
77 |
+
|
78 |
+
tracker.step(detections)
|
79 |
+
tracks = tracker.active_tracks()
|
80 |
+
|
81 |
+
for track in tracks:
|
82 |
+
x1, y1, x2, y2 = map(int, track.box)
|
83 |
+
track_id = track.id
|
84 |
+
class_id = track.class_id if track.class_id is not None else -1
|
85 |
+
class_name = COCO_CLASSES[class_id] if 0 <= class_id < len(COCO_CLASSES) else str(class_id)
|
86 |
+
|
87 |
+
cv2.rectangle(frame, (x1, y1), (x2, y2), (0, 255, 0), 2)
|
88 |
+
cv2.putText(frame, f'{class_name} {track_id}', (x1, y1 - 10),
|
89 |
+
cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 2)
|
90 |
+
|
91 |
+
out.write(frame)
|
92 |
+
|
93 |
+
cap.release()
|
94 |
+
out.release()
|
95 |
+
return output_path
|
96 |
+
|
97 |
+
def process_video(video):
|
98 |
+
output_path = track_objects(video)
|
99 |
+
return output_path
|
100 |
+
|
101 |
+
interface = gr.Interface(
|
102 |
+
fn=process_video,
|
103 |
+
inputs=gr.Video(label="Input Video"),
|
104 |
+
outputs=[
|
105 |
+
gr.Video(label="Processed Video"),
|
106 |
+
gr.File(label="Download Processed Video")
|
107 |
+
],
|
108 |
+
title="SAHI Video Object Tracker",
|
109 |
+
description="Object detection and tracking using SAHI and YOLOv11."
|
110 |
+
)
|
111 |
+
|
112 |
+
if __name__ == "__main__":
|
113 |
+
interface.launch()
|