Update detection display

app.py

@@ -3,9 +3,7 @@ import socket
 import time
 import gradio as gr
 import numpy as np
-from PIL import Image
-import supervision as sv
-import cv2
+from PIL import Image, ImageDraw
 import base64
 import requests
 import json
@@ -23,13 +21,14 @@ LINE_WIDTH = 2
 print(f"Gradio version: {gr.__version__}")
 
 # Define the inference function
-def predict_image(img, threshold):
+def predict_image(image, threshold):
     
-    if isinstance(img, Image.Image):
-        img = np.array(img)
-    
-    if not isinstance(img, np.ndarray) or len(img.shape) != 3 or img.shape[2] != 3:
-        raise BaseException("predict_image(): input 'img' shoud be single RGB image in PIL or Numpy array format.")
+    if not isinstance(image, Image.Image):
+        raise BaseException("predit_image(): input 'image' shoud be single RGB image in PIL format.")
+        
+    img = np.array(image)
+    if len(img.shape) != 3 or img.shape[2] != 3:
+        raise BaseException("predit_image(): input 'image' shoud be single RGB image in PIL format.")
 
     # Encode the image data as base64
     image_base64 = base64.b64encode(np.ascontiguousarray(img)).decode()
@@ -59,22 +58,18 @@ def predict_image(img, threshold):
     detections = json_data['detections']
     duration = json_data['duration']
     
-    # Convert the numpy array (RGB format) to a cv2 image (BGR format)
-    cv2_img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
-    
-    # Annotate the image with detections
-    oriented_box_annotator = sv.OrientedBoxAnnotator()
-    annotated_frame = oriented_box_annotator.annotate(
-        scene=cv2_img,
-        detections=detections
-    )
-    image_with_predictions_rgb = cv2.cvtColor(annotated_frame, cv2.COLOR_BGR2RGB)
+    # drow boxes on image
+    draw = ImageDraw.Draw(image)
+
+    for (class_name, coords, confidence) in detections:
+        if len(coords) != 4:
+            raise ValueError("Each detection should be a polygon with 8 coordinates (xyxyxyxy).")
 
-    img_data_in = base64.b64decode(json_data['image']) 
-    np_img = np.frombuffer(img_data_in, dtype=np.uint8).reshape(img.shape)
-    pil_img = Image.fromarray(np_img)
+        points = [(coord[0], coord[1]) for coord in coords]
+        draw.polygon(points, outline="red", width=LINE_WIDTH)
+        draw.text((points[0][0], points[0][1]), class_name, fill="red")
 
-    return pil_img, img.shape, len(detections), duration
+    return image, img.shape, len(detections), duration
 
 
 # Define example images and their true labels for users to choose from

requirements.txt

@@ -1,3 +1 @@
-opencv-python
-supervision
 requests