""" tile_counter.py Simple tile counter for images containing rectangular/square colored tiles. Install: pip install opencv-python numpy pandas Run: python tile_counter2.py "source_image2.jpg" in Windows PowerShell Output: tile_results/tile_counts.csv """ import cv2 import numpy as np import pandas as pd import sys from pathlib import Path # Approximate colors used in the sample image, in RGB. # Edit these if your real tiles use different colors. PALETTE = { "dark blue": (34, 83, 144), "light blue": (106, 144, 179), "green": (77, 110, 80), "yellow": (218, 169, 50), "red/orange": (195, 70, 43), "dark gray": (58, 57, 60), "gray": (133, 129, 127), "white": (215, 209, 203), } def nearest_color(rgb): """Return the palette color closest to an RGB value.""" rgb = np.array(rgb, dtype=float) best_name = None best_distance = float("inf") for name, color in PALETTE.items(): distance = np.linalg.norm( rgb - np.array(color, dtype=float) ) if distance < best_distance: best_distance = distance best_name = name return best_name def detect_tiles(image_path, k=9, min_area=5000): image = cv2.imread(str(image_path)) if image is None: raise FileNotFoundError( f"Could not read image: {image_path}" ) # OpenCV loads BGR. # Convert to RGB for color processing. rgb = cv2.cvtColor( image, cv2.COLOR_BGR2RGB ) # ------------------------------------------------------------ # 1. K-means color segmentation # ------------------------------------------------------------ # Reduce the image to a small version for faster clustering. small = cv2.resize( rgb, None, fx=0.25, fy=0.25 ) pixels = small.reshape( -1, 3 ).astype(np.float32) criteria = ( cv2.TERM_CRITERIA_EPS + cv2.TERM_CRITERIA_MAX_ITER, 100, 0.2, ) _, _, centers = cv2.kmeans( pixels, k, None, criteria, 10, cv2.KMEANS_PP_CENTERS, ) centers = np.uint8(centers) # Assign every full-resolution pixel # to its nearest cluster. pixels_full = rgb.reshape( -1, 3 ).astype(np.float32) distances = ( ( pixels_full[:, None, :] - centers[None, :, :] ) ** 2 ).sum(axis=2) labels = np.argmin( distances, axis=1 ) labels = labels.reshape( rgb.shape[:2] ).astype(np.uint8) # ------------------------------------------------------------ # 2. Detect tile boundaries # ------------------------------------------------------------ # The sample image has thin grout/boundary lines. # Detecting edges and removing them prevents # neighboring tiles from merging. blurred = cv2.GaussianBlur( image, (5, 5), 0 ) edges = cv2.Canny( blurred, 20, 70 ) edge_barrier = cv2.dilate( edges, np.ones((3, 3), np.uint8), iterations=1, ) detections = [] # ------------------------------------------------------------ # 3. Find connected regions for each color # ------------------------------------------------------------ for cluster_id in range(k): mask = ( labels == cluster_id ).astype(np.uint8) # Remove detected grout/edge pixels. mask[edge_barrier > 0] = 0 n_labels, component_labels, stats, centroids = ( cv2.connectedComponentsWithStats( mask, 8 ) ) for component_id in range( 1, n_labels ): x, y, w, h, area = ( stats[component_id] ) # Ignore small regions caused # by texture/noise. if area < min_area: continue # Ignore very thin regions. if w < 10 or h < 10: continue # Tile-like regions should have # a reasonably rectangular bounding box. rectangularity = ( area / float(w * h) ) if rectangularity < 0.55: continue component_mask = ( component_labels == component_id ) # Mean RGB value from the detected tile. mean_rgb = rgb[ component_mask ].mean(axis=0) color_name = nearest_color( mean_rgb ) cx, cy = centroids[ component_id ] detections.append({ "x": int(x), "y": int(y), "width": int(w), "height": int(h), "area_px": int(area), "center_x": round( float(cx), 1 ), "center_y": round( float(cy), 1 ), "mean_R": round( float(mean_rgb[0]), 1 ), "mean_G": round( float(mean_rgb[1]), 1 ), "mean_B": round( float(mean_rgb[2]), 1 ), "color": color_name, }) # Sort spatially: # top-to-bottom, then left-to-right. detections.sort( key=lambda d: ( d["center_y"], d["center_x"] ) ) # Give every tile an ID. for i, detection in enumerate( detections, start=1 ): detection["tile_id"] = i return detections def save_results(detections, output_dir): output_dir = Path(output_dir) # Create the output folder if it doesn't exist. output_dir.mkdir( parents=True, exist_ok=True ) # ------------------------------------------------------------ # CSV # ------------------------------------------------------------ df = pd.DataFrame( detections ) columns = [ "tile_id", "color", "x", "y", "width", "height", "area_px", "center_x", "center_y", "mean_R", "mean_G", "mean_B", ] df = pd.DataFrame( detections, columns=columns ) csv_path = ( output_dir / "tile_counts2.csv" ) df.to_csv( csv_path, index=False ) # ------------------------------------------------------------ # Summary printed to terminal # ------------------------------------------------------------ print() print("Detected tiles:") print("----------------") counts = ( df["color"] .value_counts() ) for color, count in counts.items(): print( f"{color:12s}: {count}" ) print() print( f"TOTAL : {len(df)}" ) return csv_path def main(): # ------------------------------------------------------------ # Check that an image was provided. # ------------------------------------------------------------ if len(sys.argv) < 2: print("Usage:") print( " python tile_counter.py image.jpg" ) return # Get image path from command line. image_path = Path( sys.argv[1] ) # Check that the file exists. if not image_path.exists(): print( f"ERROR: Image not found:" ) print( f" {image_path}" ) return print() print("Processing image...") print( f"Input: {image_path}" ) # ------------------------------------------------------------ # Detect tiles. # ------------------------------------------------------------ detections = detect_tiles( image_path, k=9, min_area=100, ) # ------------------------------------------------------------ # Save CSV. # ------------------------------------------------------------ csv_path = save_results( detections, output_dir="tile_results", ) print() print("Saved:") print( f" {csv_path.resolve()}" ) if __name__ == "__main__": main()