Add debugs on A* path if they may get blocked in routing nodes. Add function to allow offfset of 3 segment classes

app.py

@@ -85,6 +85,10 @@ ade_palette = np.array([
     [184, 255, 0], [0, 133, 255], [255, 214, 0], [25, 194, 194], [102, 255, 0],
     [92, 0, 255]
 ], dtype=np.uint8)
+if ade_palette.shape[0] < 150: # Some update require 150 class now but we only afford 146, allow offset
+    missing = 150 - ade_palette.shape[0]
+    padding = np.zeros((missing, 3), dtype=np.uint8)
+    ade_palette = np.vstack([ade_palette, padding])
 
 custom_class_map = {
     "Garbage":                 [(255, 8, 41), (235, 255, 7), (255, 5, 153), (255, 0, 102)],
@@ -183,28 +187,39 @@ def highlight_water_mask_on_frame(frame, binary_mask, color=(255, 0, 0), alpha=0
 def astar(start, goal, occ):
     h   = lambda a,b: abs(a[0]-b[0])+abs(a[1]-b[1])
     N8  = [(-1,-1),(-1,0),(-1,1),(0,-1),(0,1),(1,-1),(1,0),(1,1)]
-    openq=[(0,start)]; g={start:0}; came={}
+    openq=[(0,start)]; g={start:0}; came={};  visited = set()
     while openq:
         _,cur=heapq.heappop(openq)
         if cur==goal:
             p=[cur];                   # reconstruct
             while cur in came: cur=came[cur]; p.append(cur)
             return p[::-1]
-        for dx,dy in N8:
-            nx,ny=cur[0]+dx,cur[1]+dy
-            # out-of-bounds / blocked
-            if not (0<=nx<640 and 0<=ny<640) or occ[ny,nx]==0: continue
-            # if diagonal, ensure both orthogonals are free
-            if abs(dx)==1 and abs(dy)==1:
-                if occ[cur[1]+dy, cur[0]]==0 or occ[cur[1], cur[0]+dx]==0:
+        if cur in visited:
+            continue
+        visited.add(cur)
+        for dx, dy in N8:
+            nx, ny = cur[0] + dx, cur[1] + dy
+            if not (0 <= nx < 640 and 0 <= ny < 640) or occ[ny, nx] == 0:
+                continue
+            if abs(dx) == 1 and abs(dy) == 1:
+                if occ[cur[1]+dy, cur[0]] == 0 or occ[cur[1], cur[0]+dx] == 0:
                     continue
-            ng=g[cur]+1
-            if (nx,ny) not in g or ng<g[(nx,ny)]:
-                g[(nx,ny)]=ng
-                f=ng+h((nx,ny),goal)
-                heapq.heappush(openq,(f,(nx,ny)))
-                came[(nx,ny)]=cur
-    return []
+            neighbor = (nx, ny)
+            ng = g[cur] + 1
+            if neighbor not in g or ng < g[neighbor]:
+                g[neighbor] = ng
+                f = ng + h(neighbor, goal)
+                heapq.heappush(openq, (f, neighbor))
+                came[neighbor] = cur
+    # Save visited search as debug image
+    visited_img = np.zeros_like(occ, dtype=np.uint8)
+    for x, y in visited:
+        visited_img[y, x] = 127
+    cv2.circle(visited_img, start[::-1], 3, 255, -1)
+    cv2.circle(visited_img, goal[::-1], 3, 255, -1)
+    cv2.imwrite("/home/user/app/outputs/debug_astar_failure.png", visited_img * 2)
+    print(f"🧨 A* failed from {start} to {goal} — frontier saved to debug_astar_failure.png")
+    return [] 
 # KNN fit optimal path
 def knn_path(start, targets, occ):
     todo = targets[:]; path=[]
@@ -429,7 +444,9 @@ def _pipeline(uid,img_path):
     seg = cv2.resize(seg_tensor.numpy(), (640, 640), interpolation=cv2.INTER_NEAREST)
     print(f"🧪 [{uid}] segmentation input shape: {inputs['pixel_values'].shape}")
     water_mask, garbage_mask, movable_mask = build_masks(seg) # movable zone = water and garbage masks
-    cv2.imwrite(f"{OUTPUT_DIR}/{uid}_movable_mask.png", movable_mask * 255)
+    for cx, cy in centres:
+        cv2.circle(movable_mask, (cx, cy), 3, 127, -1)  # gray center dots
+    cv2.imwrite(f"{OUTPUT_DIR}/{uid}_movable_with_centres.png", movable_mask * 255)
     print(f"🧩 Saved debug movable_mask: {OUTPUT_DIR}/{uid}_movable_mask.png")
 
     # 2- Garbage detection (3 models) → keep centres on water