Update app.py

app.py

@@ -66,6 +66,7 @@ def yolov8_inference(
     area_image = image.width*image.height
     object_predictions = []
     html_bot_voice = ""
+    most_close = 0
     if boxes is not None:
         det_ind = 0
         for xyxy, conf, cls in zip(boxes.xyxy, boxes.conf, boxes.cls):
@@ -81,17 +82,20 @@ def yolov8_inference(
             )
             object_predictions.append(object_prediction)
             det_ind += 1
-            if area_rate >= area_thres:
-                html_bot_voice = tts(defaul_bot_voice, language="ja")
+            if area_rate >= most_close:
+                out_img = image.crop(tuple(box)).resize((image.width, image.height))
+                most_close = area_rate
+                if area_rate >= area_thres:
+                    html_bot_voice = tts(defaul_bot_voice, language="ja")
                 
-    result = visualize_object_predictions(
-        image=np_image,
-        object_prediction_list=object_predictions,
-        rect_th=2,
-        text_th=2,
-    )
-    
-    return Image.fromarray(result["image"]), html_bot_voice
+    # result = visualize_object_predictions(
+    #     image=np_image,
+    #     object_prediction_list=object_predictions,
+    #     rect_th=2,
+    #     text_th=2,
+    # )
+    # return Image.fromarray(result["image"]), html_bot_voice
+    return out_img, html_bot_voice
 
 
 outputs = [gr.Image(type="filepath", label="Robot View"),