Update app.py
Browse files
app.py
CHANGED
@@ -5,17 +5,103 @@ os.system('git clone https://github.com/WongKinYiu/yolov7.git')
|
|
5 |
|
6 |
|
7 |
def detect(inp):
|
8 |
-
os.system('python ./yolov7/detect.py --weights
|
9 |
otp=inp.split('/')[2]
|
10 |
-
return f"./yolov7/runs/detect/
|
11 |
|
12 |
#f"./yolov7/runs/detect/exp/{otp}"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
|
14 |
inp = gr.inputs.Image(type="filepath", label="Input")
|
15 |
-
|
|
|
16 |
#.outputs.Textbox()
|
17 |
|
18 |
-
io=gr.Interface(fn=
|
19 |
#,examples=["Examples/img-300_jpg.rf.6b7b035dff1cda092ce3dc22be8d0135.jpg"]
|
20 |
io.launch(debug=True,share=False)
|
21 |
|
|
|
5 |
|
6 |
|
7 |
def detect(inp):
|
8 |
+
os.system('python ./yolov7/detect.py --weights best.pt --conf 0.25 --img-size 640 --source f{inp} "--project","./yolov7/runs/detect ')
|
9 |
otp=inp.split('/')[2]
|
10 |
+
return f"./yolov7/runs/detect/exp/*"
|
11 |
|
12 |
#f"./yolov7/runs/detect/exp/{otp}"
|
13 |
+
|
14 |
+
|
15 |
+
opt = {
|
16 |
+
|
17 |
+
"weights": "best.pt", # Path to weights file default weights are for nano model
|
18 |
+
"yaml" : "custom.yaml",
|
19 |
+
"img-size": 640, # default image size
|
20 |
+
"conf-thres": 0.25, # confidence threshold for inference.
|
21 |
+
"iou-thres" : 0.45, # NMS IoU threshold for inference.
|
22 |
+
"device" : '0', # device to run our model i.e. 0 or 0,1,2,3 or cpu
|
23 |
+
"classes" : classes_to_filter # list of classes to filter or None
|
24 |
+
|
25 |
+
}
|
26 |
+
|
27 |
+
def detect2(inp):
|
28 |
+
with torch.no_grad():
|
29 |
+
weights, imgsz = opt['weights'], opt['img-size']
|
30 |
+
set_logging()
|
31 |
+
device = select_device(opt['device'])
|
32 |
+
half = device.type != 'cpu'
|
33 |
+
model = attempt_load(weights, map_location=device) # load FP32 model
|
34 |
+
stride = int(model.stride.max()) # model stride
|
35 |
+
imgsz = check_img_size(imgsz, s=stride) # check img_size
|
36 |
+
if half:
|
37 |
+
model.half()
|
38 |
+
|
39 |
+
names = model.module.names if hasattr(model, 'module') else model.names
|
40 |
+
colors = [[random.randint(0, 255) for _ in range(3)] for _ in names]
|
41 |
+
if device.type != 'cpu':
|
42 |
+
model(torch.zeros(1, 3, imgsz, imgsz).to(device).type_as(next(model.parameters())))
|
43 |
+
|
44 |
+
img0 = cv2.imread(inp)
|
45 |
+
img = letterbox(img0, imgsz, stride=stride)[0]
|
46 |
+
img = img[:, :, ::-1].transpose(2, 0, 1) # BGR to RGB, to 3x416x416
|
47 |
+
img = np.ascontiguousarray(img)
|
48 |
+
img = torch.from_numpy(img).to(device)
|
49 |
+
img = img.half() if half else img.float() # uint8 to fp16/32
|
50 |
+
img /= 255.0 # 0 - 255 to 0.0 - 1.0
|
51 |
+
if img.ndimension() == 3:
|
52 |
+
img = img.unsqueeze(0)
|
53 |
+
|
54 |
+
# Inference
|
55 |
+
t1 = time_synchronized()
|
56 |
+
pred = model(img, augment= False)[0]
|
57 |
+
|
58 |
+
# Apply NMS
|
59 |
+
classes = None
|
60 |
+
if opt['classes']:
|
61 |
+
classes = []
|
62 |
+
for class_name in opt['classes']:
|
63 |
+
|
64 |
+
classes.append(names.index(class_name))
|
65 |
+
|
66 |
+
if classes:
|
67 |
+
|
68 |
+
classes = [i for i in range(len(names)) if i not in classes]
|
69 |
+
|
70 |
+
|
71 |
+
pred = non_max_suppression(pred, opt['conf-thres'], opt['iou-thres'], classes= [17], agnostic= False)
|
72 |
+
t2 = time_synchronized()
|
73 |
+
for i, det in enumerate(pred):
|
74 |
+
s = ''
|
75 |
+
s += '%gx%g ' % img.shape[2:] # print string
|
76 |
+
gn = torch.tensor(img0.shape)[[1, 0, 1, 0]]
|
77 |
+
if len(det):
|
78 |
+
det[:, :4] = scale_coords(img.shape[2:], det[:, :4], img0.shape).round()
|
79 |
+
|
80 |
+
for c in det[:, -1].unique():
|
81 |
+
n = (det[:, -1] == c).sum() # detections per class
|
82 |
+
s += f"{n} {names[int(c)]}{'s' * (n > 1)}, " # add to string
|
83 |
+
|
84 |
+
for *xyxy, conf, cls in reversed(det):
|
85 |
+
|
86 |
+
label = f'{names[int(cls)]} {conf:.2f}'
|
87 |
+
plot_one_box(xyxy, img0, label=label, color=colors[int(cls)], line_thickness=3)
|
88 |
+
return img0
|
89 |
+
|
90 |
+
|
91 |
+
|
92 |
+
|
93 |
+
|
94 |
+
|
95 |
+
|
96 |
+
|
97 |
+
|
98 |
|
99 |
inp = gr.inputs.Image(type="filepath", label="Input")
|
100 |
+
outputs=gr.outputs.Image(type="pil", label="Output Image")
|
101 |
+
#output = gr.outputs.Image(type="filepath", label="Output")
|
102 |
#.outputs.Textbox()
|
103 |
|
104 |
+
io=gr.Interface(fn=detect2, inputs=inp, outputs=output, title='Pot Hole Detection With Custom YOLOv7 ',examples=[["Examples/img-300_jpg.rf.6b7b035dff1cda092ce3dc22be8d0135.jpg"]])
|
105 |
#,examples=["Examples/img-300_jpg.rf.6b7b035dff1cda092ce3dc22be8d0135.jpg"]
|
106 |
io.launch(debug=True,share=False)
|
107 |
|