Update app.py

app.py

@@ -336,7 +336,6 @@
 #         debug=False
 #     )
 
-
 import os
 import io
 import base64
@@ -458,10 +457,17 @@ def encode_pil_to_dataurl(pil_img: Image.Image, fmt="PNG") -> str:
     return "data:image/{};base64,".format(fmt.lower()) + base64.b64encode(buf.getvalue()).decode("ascii")
 
 
-def annotate_segmentation(image: Image.Image, detections: sv.Detections) -> Image.Image:
+def annotate_segmentation(image: Image.Image, detections: sv.Detections, 
+                         show_labels: bool = True, show_confidence: bool = True) -> Image.Image:
     """
     Annotate image with segmentation masks using supervision library.
     This matches the visualization from rfdetr_seg_infer.py script.
+    
+    Args:
+        image: Input PIL Image
+        detections: Supervision Detections object
+        show_labels: Whether to show "Tulsi" label text
+        show_confidence: Whether to show confidence scores
     """
     try:
         # Define color palette
@@ -478,29 +484,37 @@ def annotate_segmentation(image: Image.Image, detections: sv.Detections) -> Imag
         # Create annotators
         mask_annotator = sv.MaskAnnotator(color=palette)
         polygon_annotator = sv.PolygonAnnotator(color=sv.Color.WHITE)
-        label_annotator = sv.LabelAnnotator(
-            color=palette,
-            text_color=sv.Color.BLACK,
-            text_scale=text_scale,
-            text_position=sv.Position.CENTER_OF_MASS
-        )
-
-        # Create labels with confidence scores
-        labels = [
-            f"Tulsi {float(conf):.2f}"
-            for conf in detections.confidence
-        ]
-        
-        print(f"[INFO] Annotating {len(labels)} detections")
-
-        # Apply annotations step by step
+
+        # Apply base annotations (masks and polygons always shown)
         out = image.copy()
         print("[INFO] Applying mask annotation...")
         out = mask_annotator.annotate(out, detections)
         print("[INFO] Applying polygon annotation...")
         out = polygon_annotator.annotate(out, detections)
-        print("[INFO] Applying label annotation...")
-        out = label_annotator.annotate(out, detections, labels)
+        
+        # Only add labels if at least one option is enabled
+        if show_labels or show_confidence:
+            label_annotator = sv.LabelAnnotator(
+                color=palette,
+                text_color=sv.Color.BLACK,
+                text_scale=text_scale,
+                text_position=sv.Position.CENTER_OF_MASS
+            )
+            
+            # Create labels based on options
+            labels = []
+            for conf in detections.confidence:
+                label_parts = []
+                if show_labels:
+                    label_parts.append("Tulsi")
+                if show_confidence:
+                    label_parts.append(f"{float(conf):.2f}")
+                labels.append(" ".join(label_parts))
+            
+            print(f"[INFO] Applying label annotation with {len(labels)} labels...")
+            out = label_annotator.annotate(out, detections, labels)
+        else:
+            print("[INFO] Skipping label annotation (both labels and confidence disabled)")
         
         print("[INFO] Annotation complete")
         return out
@@ -537,7 +551,7 @@ def predict():
     """
     Accepts:
       - multipart/form-data with file field "file"
-      - or JSON {"image": "<data:url...>", "conf": 0.25, "show_labels": true, "show_confidence": true}
+      - or JSON {"image": "<data:url...>", "conf": 0.05, "show_labels": true, "show_confidence": true}
     Returns JSON:
       {"annotated": "<data:image/png;base64,...>", "confidences": [..], "count": N}
     """
@@ -588,6 +602,7 @@ def predict():
         show_confidence = payload.get("show_confidence", True)
 
     print(f"[INFO] Image size: {img.size}, Confidence threshold: {conf_threshold}")
+    print(f"[INFO] Display options - Labels: {show_labels}, Confidence: {show_confidence}")
 
     # Optionally downscale large images to reduce memory usage
     MAX_SIZE = 1024
@@ -628,7 +643,7 @@ def predict():
         
         # Annotate image using supervision library
         print("[INFO] Starting annotation...")
-        annotated_pil = annotate_segmentation(img, detections)
+        annotated_pil = annotate_segmentation(img, detections, show_labels, show_confidence)
         
         # Extract confidence scores
         confidences = [float(conf) for conf in detections.confidence]