segmentation/make_final_mask.py

import cv2
import glob
from tqdm import tqdm
import numpy as np
import os

def make_final_mask(seg_path, sam_path, result_path):
    # Try multiple extensions for seg_path
    seg_patterns = [
        os.path.join(seg_path, '*.png'),
        os.path.join(seg_path, '*.jpg'),
        os.path.join(seg_path, '*.jpeg'),
    ]
    
    seg_full_path = []
    for pattern in seg_patterns:
        files = sorted(glob.glob(pattern))
        seg_full_path.extend(files)
    seg_full_path = sorted(list(set(seg_full_path)))  # Remove duplicates
    
    # Try multiple extensions for sam_path
    sam_patterns = [
        os.path.join(sam_path, '*.jpg'),
        os.path.join(sam_path, '*.png'),
        os.path.join(sam_path, '*.jpeg'),
    ]
    
    sam_full_path = []
    for pattern in sam_patterns:
        files = sorted(glob.glob(pattern))
        sam_full_path.extend(files)
    sam_full_path = sorted(list(set(sam_full_path)))
    
    # DEBUG
    print(f"[DEBUG] seg_path: {seg_path}")
    print(f"[DEBUG] Found {len(seg_full_path)} seg images")
    if len(seg_full_path) > 0:
        print(f"[DEBUG] First 3 seg files: {seg_full_path[:3]}")
    
    print(f"[DEBUG] sam_path: {sam_path}")
    print(f"[DEBUG] Found {len(sam_full_path)} sam images")
    if len(sam_full_path) > 0:
        print(f"[DEBUG] First 3 sam files: {sam_full_path[:3]}")
    
    if len(seg_full_path) == 0:
        print(f"[ERROR] No seg images found in {seg_path}")
        return
    
    if len(sam_full_path) == 0:
        print(f"[ERROR] No sam images found in {sam_path}")
        return
    
    # Match by filename (without extension)
    seg_dict = {}
    for path in seg_full_path:
        basename = os.path.splitext(os.path.basename(path))[0]
        seg_dict[basename] = path
    
    sam_dict = {}
    for path in sam_full_path:
        basename = os.path.splitext(os.path.basename(path))[0]
        sam_dict[basename] = path
    
    # Find matching pairs
    matched_pairs = []
    for name in seg_dict.keys():
        if name in sam_dict:
            matched_pairs.append((seg_dict[name], sam_dict[name]))
    
    print(f"[INFO] Found {len(matched_pairs)} matching pairs")
    
    if len(matched_pairs) == 0:
        print("[ERROR] No matching pairs found!")
        print(f"[DEBUG] Seg basenames: {list(seg_dict.keys())[:5]}")
        print(f"[DEBUG] Sam basenames: {list(sam_dict.keys())[:5]}")
        return
    
    for seg, sam in tqdm(matched_pairs):
        seg_img = cv2.imread(seg)
        sam_img = cv2.imread(sam)
        
        if seg_img is None:
            print(f"[WARN] Failed to read seg image: {seg}")
            continue
        if sam_img is None:
            print(f"[WARN] Failed to read sam image: {sam}")
            continue
        
        # Resize if shapes don't match
        if seg_img.shape != sam_img.shape:
            print(f"[INFO] Resizing sam {sam_img.shape} to match seg {seg_img.shape}")
            sam_img = cv2.resize(sam_img, (seg_img.shape[1], seg_img.shape[0]))
        
        img_name = os.path.basename(sam)
        added_img = cv2.bitwise_and(seg_img, sam_img)
        binary_map = cv2.cvtColor(added_img, cv2.COLOR_BGR2GRAY)
        
        nlabels, labels, stats, centroids = cv2.connectedComponentsWithStats(
            binary_map, None, None, None, 8, cv2.CV_32S
        )
        
        # Get CC_STAT_AREA component as stats[label, COLUMN]
        areas = stats[1:, cv2.CC_STAT_AREA]
        result = np.zeros((labels.shape), np.uint8)
        for i in range(0, nlabels - 1):
            if areas[i] >= 400:   # Keep
                result[labels == i + 1] = 255
        
        output_path = os.path.join(result_path, img_name)
        cv2.imwrite(output_path, result)
        print(f"[INFO] Saved: {output_path}")

if __name__ == '__main__':
    seg_path = '/Users/Admin/ScalpVision/datasets/seg_train'  # mask gốc
    sam_path = '/Users/Admin/ScalpVision/prediction/sam_result/sam_val'  # mask SAM
    result_path = 'prediction/ensemble_result/ensemble_val'  # output mask hợp nhất
    os.makedirs(result_path, exist_ok=True)
    make_final_mask(seg_path, sam_path, result_path)