changes

scripts/welspun_classifier.py

@@ -56,7 +56,6 @@ class Welspun_Classifier(ModelWrapper):
         self.defect_type = ""
         self.frame_skip_count = 25
         self.counter = 0
-        self.k = 0
 
     def _pre_process(self, x):
         """
@@ -189,29 +188,28 @@ class Welspun_Classifier(ModelWrapper):
         return exp_vec / np.sum(exp_vec)
 
     def process_frame(self, frame):
-        if self.k % 10 == 0:
-            self.k = self.k + 1
-            starttime = time.time()
-            vino_frame = frame.copy()
-            # vino_frame = vino_frame[20:600,150:650]
-            vino_frame = vino_frame[5: 600, 70: 650]
-            images = np.ndarray(shape=(self.n, self.c, self.h, self.w))
-            images_hw = []
-            for i in range(self.n):
-                image = vino_frame
-                ih, iw = image.shape[:-1]
-                images_hw.append((ih, iw))
+        starttime = time.time()
+        vino_frame = frame.copy()
+        # vino_frame = vino_frame[20:600,150:650]
+        vino_frame = vino_frame[5: 600, 70: 650]
+        images = np.ndarray(shape=(self.n, self.c, self.h, self.w))
+        images_hw = []
+        for i in range(self.n):
+            image = vino_frame
+            ih, iw = image.shape[:-1]
+            images_hw.append((ih, iw))
 
-                if (ih, iw) != (self.h, self.w):
-                    # log.warning("Image {} is resized from {} to {}".format(img, image.shape[:-1], (h, w)))
-                    image = cv2.resize(image, (self.w, self.h))
-                    image = self.normalize(image)
-                    # log.warning("Image {} is resized from {} to {}".format(args.input[i], image.shape[:-1], (h, w)))
-                image = image.transpose((2, 0, 1))  # Change data layout from HWC to CHW
-                images[i] = image
-            # log.info("Batch size is {}".format(n))
-            #
-            # log.info("Starting inference in synchronous mode")
+            if (ih, iw) != (self.h, self.w):
+                # log.warning("Image {} is resized from {} to {}".format(img, image.shape[:-1], (h, w)))
+                image = cv2.resize(image, (self.w, self.h))
+                image = self.normalize(image)
+                # log.warning("Image {} is resized from {} to {}".format(args.input[i], image.shape[:-1], (h, w)))
+            image = image.transpose((2, 0, 1))  # Change data layout from HWC to CHW
+            images[i] = image
+        # log.info("Batch size is {}".format(n))
+        #
+        # log.info("Starting inference in synchronous mode")
+        if self.counter % 10 == 0:
             start = time.time()
             res = self.exec_net.infer(inputs={self.input_blob: images})
             print(f"Inference time: {time.time() - start}")
@@ -293,8 +291,7 @@ class Welspun_Classifier(ModelWrapper):
                                 color=(0, 0, 255),
                                 thickness=2,
                                 fontScale=1, fontFace=cv2.LINE_AA)
-                    self.send_payload("Split Defect Detected", resized_frame, "Split " + str(prob[3]), "#472020",
-                                      "#ed2020",
+                    self.send_payload("Split Defect Detected", resized_frame, "Split " + str(prob[3]), "#472020", "#ed2020",
                                       "sound_1")
                     logger.info(f"Probability: {prob}")
                 # if self.counter % 25 == 0:
@@ -354,5 +351,6 @@ class Welspun_Classifier(ModelWrapper):
             print("total time taken to process-------------> ", str(time.time() - starttime))
             # logger.info(f"total time taken to process----------------- {time.time()-starttime}")
             # cv2.imshow('res', frame)
-
+        else:
+            pass
         return frame