Spaces:
Sleeping
Sleeping
File size: 7,062 Bytes
24cd5a5 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 | import json
import os
from pathlib import Path
import pandas as pd
from tqdm import tqdm
from collections import defaultdict
class MAdVerseDataLoaderOptimized:
def __init__(self, data_path="./data"):
self.data_path = Path(data_path)
self.annotations_path = self.data_path / "annotations"
self.images_path = self.data_path / "images"
self.file_index = {} # Fast lookup: filename -> full_path
def build_file_index(self):
"""Build complete file index once (FAST!)"""
print("\n" + "="*80)
print("Building file index (one-time scan)...")
print("="*80)
file_count = 0
for source_folder in ['Advert_Gallery', 'OnlineAds', 'Epaper1', 'Epaper2']:
search_dir = self.images_path / source_folder
if not search_dir.exists():
print(f"β οΈ Folder not found: {search_dir}")
continue
print(f"\nScanning {source_folder}...")
for root, dirs, files in os.walk(search_dir):
for filename in files:
if filename.lower().endswith(('.jpg', '.jpeg', '.png')):
full_path = str(Path(root) / filename)
self.file_index[filename.lower()] = full_path
file_count += 1
print(f" β Found {file_count} images so far")
print(f"\nβ Total files indexed: {len(self.file_index)}")
print("="*80)
def load_annotations(self):
"""Load all annotation files"""
annotations = {}
print("\n" + "="*80)
print("Loading annotations...")
print("="*80 + "\n")
for json_file in self.annotations_path.glob("*.json"):
source_name = json_file.stem.replace("_annotation", "").replace("_annot_j", "")
print(f"Loading {source_name}...", end=" ")
with open(json_file, 'r', encoding='utf-8') as f:
data = json.load(f)
annotations[source_name] = data
print(f"β ({len(data)} items)")
return annotations
def create_metadata_df(self, annotations):
"""Convert annotations to structured DataFrame"""
rows = []
print("\n" + "="*80)
print("Processing annotations...")
print("="*80 + "\n")
for source, data in annotations.items():
print(f"Processing {source} ({len(data)} items)...")
for item in tqdm(data, desc=f" {source}", ncols=80):
row = self._extract_row_from_madverse_item(item, source)
if row:
rows.append(row)
df = pd.DataFrame(rows)
print(f"\nβ Created DataFrame with {len(df)} rows")
# Fast image path lookup using pre-built index
print("\nLinking image paths (using index)...")
df['image_path'] = df['image_filename'].apply(self._fast_find_image)
# Remove rows where images don't exist
before_count = len(df)
df = df[df['image_path'].notna()]
found_count = len(df)
missing_count = before_count - found_count
print(f"β Found {found_count:,} images")
print(f"β Missing {missing_count:,} images ({missing_count/before_count*100:.1f}%)")
return df
def _extract_row_from_madverse_item(self, item, source):
"""Extract row from MAdVerse annotation format"""
# Extract filename from img_path
img_path = item.get('img_path', '')
# Handle both forward and backward slashes
img_path = img_path.replace('\\\\', '/').replace('\\', '/')
filename = Path(img_path).name
# Extract hierarchical annotation
hier_annot = item.get('hier_annot', [])
category = hier_annot[0] if len(hier_annot) > 0 else 'unknown'
subcategory = hier_annot[1] if len(hier_annot) > 1 else ''
brand = hier_annot[2] if len(hier_annot) > 2 else ''
# Map source to folder name
source_folder = self._map_source_to_folder(source)
# Create image_id from filename (without extension)
image_id = Path(filename).stem
row = {
'source': source,
'source_folder': source_folder,
'image_id': image_id,
'image_filename': filename,
'category': category,
'subcategory': subcategory,
'brand': brand,
'language': item.get('language', 'unknown'),
'ad_type': item.get('ad_type', 'unknown'),
'original_path': img_path
}
return row
def _map_source_to_folder(self, annotation_source):
"""Map annotation source to actual image folder"""
mapping = {
'adgal': 'Advert_Gallery',
'epaper1': 'Epaper1',
'epaper2': 'Epaper2',
'web': 'OnlineAds'
}
return mapping.get(annotation_source, annotation_source)
def _fast_find_image(self, filename):
"""Fast O(1) lookup using pre-built index"""
return self.file_index.get(filename.lower())
def save_metadata(self, df, output_path="./processed/metadata/madverse_metadata.csv"):
"""Save metadata to CSV"""
output_path = Path(output_path)
output_path.parent.mkdir(parents=True, exist_ok=True)
df.to_csv(output_path, index=False)
print("\n" + "="*80)
print("RESULTS")
print("="*80)
print(f"\nβ Metadata saved to: {output_path}")
print(f"β Total images processed: {len(df):,}")
if len(df) > 0:
print(f"\nπ Breakdown by source:")
print(df['source'].value_counts())
print(f"\nπ Breakdown by category (top 10):")
print(df['category'].value_counts().head(10))
print(f"\nπ Breakdown by language:")
print(df['language'].value_counts())
print(f"\nπ Breakdown by ad_type:")
print(df['ad_type'].value_counts())
return df
# Main execution
if __name__ == "__main__":
import time
start_time = time.time()
print("\n" + "="*80)
print("MAdVerse Dataset Loader (OPTIMIZED)")
print("="*80)
loader = MAdVerseDataLoaderOptimized()
# Step 1: Build file index (fast one-time scan)
loader.build_file_index()
# Step 2: Load annotations
annotations = loader.load_annotations()
# Step 3: Create metadata DataFrame
df = loader.create_metadata_df(annotations)
# Step 4: Save metadata
if len(df) > 0:
loader.save_metadata(df)
print(f"\nπ Sample metadata:")
print(df[['source', 'category', 'brand', 'language', 'image_filename']].head(15))
print(f"\nβ
Columns: {list(df.columns)}")
print(f"β
Shape: {df.shape}")
else:
print("\nβ οΈ No images found. Check your image folders.")
elapsed_time = time.time() - start_time
print(f"\nβ±οΈ Total execution time: {elapsed_time:.2f} seconds ({elapsed_time/60:.2f} minutes)")
print("\n" + "="*80) |