rbg-bgr

app.py

@@ -2,6 +2,7 @@ from yolox_onnx import YOLOX_ONNX
 import albumentations as A
 import gradio as gr
 import cv2
+import numpy as np
 
 def show_example(path):
     return cv2.cvtColor(cv2.imread(path), cv2.COLOR_BGR2RGB)
@@ -27,8 +28,10 @@ def get_response(input_img,add_snow,add_rain,add_fog,confidence_threshold,iou_th
     if not hasattr(input_img,'shape'):
         return "invalid input",input_img
 
+    input_img=cv2.cvtColor(input_img,cv2.COLOR_RGB2BGR).astype(np.float32)
+
     pedestrian_detector.predict(input_img,confidence_threshold,iou_threshold)
-    out_img=pedestrian_detector.output_img
+    out_img=pedestrian_detector.output_img.astype(np.uint8)
 
     if add_snow:
         out_img = weather_transform["add_snow"] (image=out_img)['image']
@@ -40,7 +43,7 @@ def get_response(input_img,add_snow,add_rain,add_fog,confidence_threshold,iou_th
         pass
 
 
-    return out_img
+    return cv2.cvtColor(out_img,cv2.COLOR_BGR2RGB)
 
 
 if __name__ == "__main__":
@@ -59,7 +62,7 @@ if __name__ == "__main__":
         gr.Checkbox(label="add_rain"),
         gr.Checkbox(label="add_fog"),
         gr.Slider(0, 1,value=0.5, step=0.01, label="confidence_threshold"),
-        gr.Slider(0, 1,value=0.45, step=0.01, label="iou_threshold")],
+        gr.Slider(0, 1,value=0.4, step=0.01, label="iou_threshold")],
         examples=[[show_example('test-images/test1.jpg')],[show_example('test-images/test2.jpg')],[show_example('test-images/test3.jpg')]],
         outputs=[gr.Image(type="numpy")],
         title="Pedestrian Detection with All weather augmentation",

yolox_onnx.py

@@ -2,7 +2,6 @@ import cv2
 from time import time
 import numpy as np
 import onnxruntime
-from collections import defaultdict
 
 
 class YOLOX_ONNX: