Commit @6e76cad9442bcff0c1da5a92c1f42df9dfea3a68

894e2c8

6e76cad

hls_streaming/hls.py

--- hls_streaming/hls.py

+++ hls_streaming/hls.py


                             self.send_image_to_server(img_binary)
                             cv2.imwrite(f'hls_streaming/captured_frame_/{self.cctvid}_{datetime.now()}_{frame_name}', self.current_frame)
                             self.last_capture_time = current_time
-                            print(f"Captured {frame_name} of {self.cctvid} at time: {current_time - self.start_time:.2f}s")
+                            # print(f"Captured {frame_name} of {self.cctvid} at time: {current_time - self.start_time:.2f}s")
                             self.captured_frame_count +=1
 
             time.sleep(0.1)

894e2c8

6e76cad

inference_endpoint.py

--- inference_endpoint.py

+++ inference_endpoint.py


 
         if len(mask_maps) != 0:
             seg_image = overlay_mask(image, mask_maps[0], color=(0, 255, 0), alpha=0.3)
+            area_percent = 0
+        else :
+            area_percent = np.sum(mask_maps) / image.shape[0] * image.shape[1]
 
+        # write another post request for pushing a detection result
         return {"message": f"Image {filename} uploaded successfully!"}
 
+    def send_result(self):
+        pass
+
 if __name__ == '__main__':
     app.run(debug=True, port=12345)

6e76cad

postprocessing.py (added)

+++ postprocessing.py

...	...	@@ -0,0 +1,17 @@
	1	+import numpy as np
	2	+from flask import Flask, request
	3	+from flask_restx import Api, Resource, fields
	4	+import os
	5	+from datetime import datetime
	6	+from yoloseg.inference_ import Inference, overlay_mask
	7	+import cv2
	8	+import time
	9	+import base64
	10	+
	11	+
	12	+app = Flask(__name__)
	13	+api = Api(app, version='1.0', title='Detection postprocessing component',
	14	+ description='A postprocessing that dynamically decides how frequently send image and detection result')
	15	+
	16	+# Namespace definition
	17	+ns = api.namespace('cctv', description='CCTV operations')(파일 끝에 줄바꿈 문자 없음)

894e2c8

6e76cad

run_image_anal_backend.sh

--- run_image_anal_backend.sh

+++ run_image_anal_backend.sh


 python streaming_process.py --cctv_num 1 &
 pids+=($!)
 
-python test.py
+
+python inference_endpoint.py
+#gunicorn --workers=6 inference_endpoint:app
 pids+=($!)
 
 # Function to kill all processes

894e2c8

6e76cad

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
             if confidence > thres:
                 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
+                if w < w_thres or h < h_thres:
+                    continue
                 left = int((x - 0.5 * w) * x_factor)
                 top = int((y - 0.5 * h) * y_factor)
                 width = int(w * x_factor)

         for idx, det in enumerate(detections):
             box = det['box']
             x1, y1, w, h = box
+            print(f"x1 : {x1}, y1 : {y1}, w: {w}, h: {h}")
+
+            x1, y1, x2, y2 = x1, y1, x1 + w, y1 + h
 
             #... why the model outputs ... negative values?...
             if x1 <= 0 :
+                w += x1
                 x1 = 0
             if y1 <= 0 :
+                h += y1
                 y1 = 0
-            x1, y1, x2, y2 = x1, y1, x1 + w, y1 + h
 
-            # To handle edge cases where you get bboxes that pass beyond the original image_binary
+            # To handle edge cases where you get bboxes that pass beyond the original size of image_binary
             if y2 > image_shape[1]:
-                h = h + image_shape[1] - h - y1
+                h = image_shape[1] - y1
             if x2 > image_shape[0]:
-                w = w + image_shape[1] - w - y1
+                w = image_shape[1] - y1
+
+            print(f"x2: {x2}, y2 : {y2}")
 
             # Get the corresponding mask coefficients for this detection
             coeffs = det["mask_coefficients"]

...	...	@@ -69,8 +69,15 @@
69	69
70	70	if len(mask_maps) != 0:
71	71	seg_image = overlay_mask(image, mask_maps[0], color=(0, 255, 0), alpha=0.3)
	72	+ area_percent = 0
	73	+ else :
	74	+ area_percent = np.sum(mask_maps) / image.shape[0] * image.shape[1]
72	75
	76	+ # write another post request for pushing a detection result
73	77	return {"message": f"Image {filename} uploaded successfully!"}
74	78
	79	+ def send_result(self):
	80	+ pass
	81	+
75	82	if __name__ == '__main__':
76	83	app.run(debug=True, port=12345)

...	...	@@ -58,8 +58,13 @@
58	58	confidence = scores_classification[class_id]
59	59
60	60	thres = self.model_score_threshold
	61	+ w_thres = 20
	62	+ h_thres = 20
61	63	if confidence > thres:
62	64	x, y, w, h = detection[:4]
	65	+ # if bboxes are too small, it just skips, and it is not a bad idea since we do not need to detect small areas
	66	+ if w < w_thres or h < h_thres:
	67	+ continue
63	68	left = int((x - 0.5 * w) * x_factor)
64	69	top = int((y - 0.5 * h) * y_factor)
65	70	width = int(w * x_factor)
...	...	@@ -97,19 +102,25 @@
97	102	for idx, det in enumerate(detections):
98	103	box = det['box']
99	104	x1, y1, w, h = box
	105	+ print(f"x1 : {x1}, y1 : {y1}, w: {w}, h: {h}")
	106	+
	107	+ x1, y1, x2, y2 = x1, y1, x1 + w, y1 + h
100	108
101	109	#... why the model outputs ... negative values?...
102	110	if x1 <= 0 :
	111	+ w += x1
103	112	x1 = 0
104	113	if y1 <= 0 :
	114	+ h += y1
105	115	y1 = 0
106		- x1, y1, x2, y2 = x1, y1, x1 + w, y1 + h
107	116
108		- # To handle edge cases where you get bboxes that pass beyond the original image_binary
	117	+ # To handle edge cases where you get bboxes that pass beyond the original size of image_binary
109	118	if y2 > image_shape[1]:
110		- h = h + image_shape[1] - h - y1
	119	+ h = image_shape[1] - y1
111	120	if x2 > image_shape[0]:
112		- w = w + image_shape[1] - w - y1
	121	+ w = image_shape[1] - y1
	122	+
	123	+ print(f"x2: {x2}, y2 : {y2}")
113	124
114	125	# Get the corresponding mask coefficients for this detection
115	126	coeffs = det["mask_coefficients"]

...	...	@@ -90,7 +90,7 @@
90	90	self.send_image_to_server(img_binary)
91	91	cv2.imwrite(f'hls_streaming/captured_frame_/{self.cctvid}_{datetime.now()}_{frame_name}', self.current_frame)
92	92	self.last_capture_time = current_time
93		- print(f"Captured {frame_name} of {self.cctvid} at time: {current_time - self.start_time:.2f}s")
	93	+ # print(f"Captured {frame_name} of {self.cctvid} at time: {current_time - self.start_time:.2f}s")
94	94	self.captured_frame_count +=1
95	95
96	96	time.sleep(0.1)

...	...	@@ -11,7 +11,9 @@
11	11	python streaming_process.py --cctv_num 1 &
12	12	pids+=($!)
13	13
14		-python test.py
	14	+
	15	+python inference_endpoint.py
	16	+#gunicorn --workers=6 inference_endpoint:app
15	17	pids+=($!)
16	18
17	19	# Function to kill all processes

Delete comment