Distinguish ambigouity mapping'

app.py

@@ -84,7 +84,7 @@ ade_palette = np.array([
 ], dtype=np.uint8)
 
 custom_class_map = {
-    "Garbage":                 [(150, 5, 61)],
+    "Garbage":                 [(255, 8, 41)],
     "Water":                   [(0, 102, 200), (11, 102, 255), (31, 0, 255)],
     "Grass / Vegetation":      [(10, 255, 71), (143, 255, 140)],
     "Tree / Natural Obstacle": [(4, 200, 3), (235, 12, 255), (255, 6, 82), (255, 163, 0)],
@@ -96,6 +96,13 @@ custom_class_map = {
 }
 TOL = 30  # RGB tolerance
 
+# Segment class [150, 5, 61] is only detectable as garbage if it's large enough 
+def interpret_rgb_class(decoded_img):
+    ambiguous_rgb = np.array([150, 5, 61])
+    matches = np.all(np.abs(decoded_img - ambiguous_rgb) <= TOL, axis=-1)
+    match_ratio = np.count_nonzero(matches) / matches.size
+    return "garbage" if match_ratio > 0.15 else "sand"
+
 # Masking zones (Garbage and Water zone to be travelable)
 def build_masks(seg):
     """
@@ -107,6 +114,16 @@ def build_masks(seg):
     decoded = ade_palette[seg]
     water_mask   = np.zeros(seg.shape, np.uint8)
     garbage_mask = np.zeros_like(water_mask)
+    # Resolve ambiguity: (150,5,61) → Sand or Garbage?
+    context_label = interpret_rgb_class(decoded)
+    resolved_map = custom_class_map.copy()
+    # Dynamically re-assign the ambiguous RGB class
+    if context_label == "garbage":
+        resolved_map["Garbage"].append((150, 5, 61))
+        resolved_map["Sand / Soil / Ground"] = [rgb for rgb in resolved_map["Sand / Soil / Ground"] if rgb != (150, 5, 61)]
+    else: # Fall back as appointed to be sth else
+        resolved_map["Sand / Soil / Ground"].append((150, 5, 61))
+        resolved_map["Garbage"] = [rgb for rgb in resolved_map["Garbage"] if rgb != (150, 5, 61)]
     # Append water pixels to water_mask
     for rgb in custom_class_map["Water"]:
         water_mask |= (np.abs(decoded - rgb).max(axis=-1) <= TOL)
@@ -117,7 +134,7 @@ def build_masks(seg):
     return water_mask, garbage_mask, movable_mask
 
 # Garbage mask can be highlighted in red
-def highlight_chunk_masks_on_frame(frame, labels, objs, color_uncollected=(0, 0, 128), color_collected=(0, 128, 0), alpha=0.3):
+def highlight_chunk_masks_on_frame(frame, labels, objs, color_uncollected=(0, 0, 128), color_collected=(0, 128, 0), alpha=0.6):
     """
     Overlays semi-transparent colored regions for garbage chunks on the frame.
     `objs` must have 'pos' and 'col' keys. The collection status changes the overlay color.
@@ -175,7 +192,7 @@ def knn_path(start, targets, occ):
 
 # ── Robot sprite/class -──────────────────────────────────────────────────
 class Robot:
-    def __init__(self, sprite, speed=200): # Declare the robot's physical stats and routing (position, speed, movement, path)
+    def __init__(self, sprite, speed=2000): # Declare the robot's physical stats and routing (position, speed, movement, path)
         self.png = np.array(Image.open(sprite).convert("RGBA").resize((40,40)))
         self.pos = [0,0]; self.speed=speed
     def step(self, path):