Commit @dd791b5cd9c9458c00fa6fb59e296e152f32c1fe

윤영준 05-23

inference result sending fix

exception handling when there is no mask

@dd791b5cd9c9458c00fa6fb59e296e152f32c1fe

936fe0e

dd791b5

inference_endpoint.py

--- inference_endpoint.py

+++ inference_endpoint.py


 from yoloseg.inference_ import Inference, overlay_mask
 import cv2
 import time
+from zoneinfo import ZoneInfo
 import base64
 import requests
 from requests_toolbelt import MultipartEncoder

         self.image_type = None
         self.seg_image = None
         self.area_percent = 0
+        self.time_zone = self.time_zone = ZoneInfo("Asia/Seoul")
 
     @ns.response(200, 'Success')
     @ns.response(400, 'Validation Error')

 
         t2 = time.time()
         if len(self.mask) > 0:
+            self.mask = cv2.resize(self.mask, (image.shape[0], image.shape[1]))
             self.mask_blob = cv2.imencode('.png', self.mask)
             self.mask_blob = self.mask.tobytes()
-            self.mask = cv2.resize(self.mask, (image.shape[0], image.shape[1]))
+
 
         print(t2 - t1)
 

         else :
             self.area_percent = np.sum(self.mask) / image.shape[0] * image.shape[1]
 
-        # self.send_result()
+        self.send_result()
         # write another post request for pushing a detection result
         return {"message": f"Image {self.mask} uploaded successfully!"}
 

             'x-area-percentage' : str(self.area_percent),
         }
         session = requests.Session()
-
+        seg_binary = cv2.imencode('.png', self.seg_image)
+        seg_binary = seg_binary[1].tobytes()
+        print(type(self.seg_image))
+        print(type(self.mask))
+        print(type(self.image))
         try:
-            multipart_data = MultipartEncoder(
-                fields={
-                    'image': (
-                        f'frame_{self.cctv_name}.{self.image_type}',
-                        self.image,
-                        f'image/{self.image_type}'
-                    ),
-                    'mask' : (
-                        f'frame_mask_{self.cctv_name}.{self.image_type}',
-                        self.mask_blob,
-                        f'image/{self.image_type}'
-                    ),
-                    'seg_mask' : (
-                        f'frame_seg_{self.cctv_name}.{self.image_type}',
-                        self.seg_image,
-                        f'image/{self.image_type}'
-                    )
-                }
-            )
+            if len(self.mask) != 0:
+                multipart_data = MultipartEncoder(
+                    fields={
+                        'image': (
+                            f'frame_{self.cctv_name}.{self.image_type}',
+                            self.image,
+                            f'image/{self.image_type}'
+                        ),
+                        'mask' : (
+                            f'frame_mask_{self.cctv_name}.{self.image_type}',
+                            self.mask_blob,
+                            f'image/{self.image_type}'
+                        ),
+                        'seg_mask' : (
+                            f'frame_seg_{self.cctv_name}.{self.image_type}',
+                            seg_binary,
+                            f'image/{self.image_type}'
+                        )
+                    }
+                )
+                header["Content-Type"] = multipart_data.content_type
+                response = session.post(self.endpoint, headers=header, data=multipart_data)
+            else:
+                multipart_data = MultipartEncoder(
+                    fields={
+                        'image': (
+                            f'frame_{self.cctv_name}.{self.image_type}',
+                            self.image,
+                            f'image/{self.image_type}'
+                        ),
+                    }
+                )
             header["Content-Type"] = multipart_data.content_type
             response = session.post(self.endpoint, headers=header, data=multipart_data)
 

936fe0e

dd791b5

yoloseg/inference_.py

--- yoloseg/inference_.py

+++ yoloseg/inference_.py


             confidence = scores_classification[class_id]
 
             thres = self.model_score_threshold
-            w_thres = 20
-            h_thres = 20
+            w_thres = 40
+            h_thres = 40
 
             x, y, w, h = detection[:4]
             # if bboxes are too small, it just skips, and it is not a bad idea since we do not need to detect small areas

Add a comment

Open 0
Closed 0

List

...	...	@@ -6,6 +6,7 @@
6	6	from yoloseg.inference_ import Inference, overlay_mask
7	7	import cv2
8	8	import time
	9	+from zoneinfo import ZoneInfo
9	10	import base64
10	11	import requests
11	12	from requests_toolbelt import MultipartEncoder
...	...	@@ -59,6 +60,7 @@
59	60	self.image_type = None
60	61	self.seg_image = None
61	62	self.area_percent = 0
	63	+ self.time_zone = self.time_zone = ZoneInfo("Asia/Seoul")
62	64
63	65	@ns.response(200, 'Success')
64	66	@ns.response(400, 'Validation Error')
...	...	@@ -83,9 +85,10 @@
83	85
84	86	t2 = time.time()
85	87	if len(self.mask) > 0:
	88	+ self.mask = cv2.resize(self.mask, (image.shape[0], image.shape[1]))
86	89	self.mask_blob = cv2.imencode('.png', self.mask)
87	90	self.mask_blob = self.mask.tobytes()
88		- self.mask = cv2.resize(self.mask, (image.shape[0], image.shape[1]))
	91	+
89	92
90	93	print(t2 - t1)
91	94
...	...	@@ -95,7 +98,7 @@
95	98	else :
96	99	self.area_percent = np.sum(self.mask) / image.shape[0] * image.shape[1]
97	100
98		- # self.send_result()
	101	+ self.send_result()
99	102	# write another post request for pushing a detection result
100	103	return {"message": f"Image {self.mask} uploaded successfully!"}
101	104
...	...	@@ -110,27 +113,44 @@
110	113	'x-area-percentage' : str(self.area_percent),
111	114	}
112	115	session = requests.Session()
113		-
	116	+ seg_binary = cv2.imencode('.png', self.seg_image)
	117	+ seg_binary = seg_binary[1].tobytes()
	118	+ print(type(self.seg_image))
	119	+ print(type(self.mask))
	120	+ print(type(self.image))
114	121	try:
115		- multipart_data = MultipartEncoder(
116		- fields={
117		- 'image': (
118		- f'frame_{self.cctv_name}.{self.image_type}',
119		- self.image,
120		- f'image/{self.image_type}'
121		- ),
122		- 'mask' : (
123		- f'frame_mask_{self.cctv_name}.{self.image_type}',
124		- self.mask_blob,
125		- f'image/{self.image_type}'
126		- ),
127		- 'seg_mask' : (
128		- f'frame_seg_{self.cctv_name}.{self.image_type}',
129		- self.seg_image,
130		- f'image/{self.image_type}'
131		- )
132		- }
133		- )
	122	+ if len(self.mask) != 0:
	123	+ multipart_data = MultipartEncoder(
	124	+ fields={
	125	+ 'image': (
	126	+ f'frame_{self.cctv_name}.{self.image_type}',
	127	+ self.image,
	128	+ f'image/{self.image_type}'
	129	+ ),
	130	+ 'mask' : (
	131	+ f'frame_mask_{self.cctv_name}.{self.image_type}',
	132	+ self.mask_blob,
	133	+ f'image/{self.image_type}'
	134	+ ),
	135	+ 'seg_mask' : (
	136	+ f'frame_seg_{self.cctv_name}.{self.image_type}',
	137	+ seg_binary,
	138	+ f'image/{self.image_type}'
	139	+ )
	140	+ }
	141	+ )
	142	+ header["Content-Type"] = multipart_data.content_type
	143	+ response = session.post(self.endpoint, headers=header, data=multipart_data)
	144	+ else:
	145	+ multipart_data = MultipartEncoder(
	146	+ fields={
	147	+ 'image': (
	148	+ f'frame_{self.cctv_name}.{self.image_type}',
	149	+ self.image,
	150	+ f'image/{self.image_type}'
	151	+ ),
	152	+ }
	153	+ )
134	154	header["Content-Type"] = multipart_data.content_type
135	155	response = session.post(self.endpoint, headers=header, data=multipart_data)
136	156

...	...	@@ -59,8 +59,8 @@
59	59	confidence = scores_classification[class_id]
60	60
61	61	thres = self.model_score_threshold
62		- w_thres = 20
63		- h_thres = 20
	62	+ w_thres = 40
	63	+ h_thres = 40
64	64
65	65	x, y, w, h = detection[:4]
66	66	# if bboxes are too small, it just skips, and it is not a bad idea since we do not need to detect small areas

Delete comment