Spaces:
Running
Running
File size: 8,281 Bytes
206d8b5 12d831f 206d8b5 12d831f 206d8b5 12d831f 206d8b5 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 | """
scripts/download_data.py
Download TrashNet + TACO datasets and remap to 5 waste categories.
Usage:
python scripts/download_data.py --output_dir data/processed
TrashNet source: https://github.com/garythung/trashnet
TACO source: http://tacodataset.org
"""
import argparse
import logging
import os
import random
import shutil
from pathlib import Path
from PIL import Image, UnidentifiedImageError
logging.basicConfig(level=logging.INFO, format="%(levelname)s %(message)s")
log = logging.getLogger(__name__)
TRASHNET_MAP = {
"plastic": "plastic",
"paper": "paper",
"cardboard": "paper",
"metal": "metal",
"glass": "glass",
"trash": None,
}
TACO_MAP = {
"food": "organic",
"food_waste": "organic",
"vegetable": "organic",
"fruit": "organic",
"organic": "organic",
}
REALWASTE_MAP = {
"cardboard": "paper",
"food_organics": "organic",
"glass": "glass",
"metal": "metal",
"paper": "paper",
"plastic": "plastic",
"vegetation": "organic",
}
FEEDBACK_MAP = {
"plastic": "plastic",
"paper": "paper",
"organic": "organic",
"metal": "metal",
"glass": "glass",
}
LOCAL_BOOST_MAP = FEEDBACK_MAP.copy()
TARGET_CLASSES = ["plastic", "paper", "organic", "metal", "glass"]
SPLITS = {"train": 0.70, "val": 0.15, "test": 0.15}
INPUT_EXTS = {".jpg", ".jpeg", ".png", ".webp"}
def phash(path: str, size: int = 16) -> str:
"""Perceptual hash for duplicate detection."""
try:
img = Image.open(path).convert("L").resize((size, size), Image.LANCZOS)
pixels = list(img.getdata())
avg = sum(pixels) / len(pixels)
return "".join("1" if p > avg else "0" for p in pixels)
except Exception:
return ""
def verify_image(path: str) -> bool:
try:
with Image.open(path) as img:
img.verify()
return True
except (UnidentifiedImageError, Exception):
return False
def collect_images(source_dir: str, class_map: dict) -> dict:
"""Walk source_dir and return {target_class: [abs_path, ...]}."""
collected = {c: [] for c in TARGET_CLASSES}
for folder in Path(source_dir).iterdir():
if not folder.is_dir():
continue
target = class_map.get(folder.name.lower())
if target is None:
continue
for file_path in folder.rglob("*"):
if file_path.suffix.lower() in INPUT_EXTS:
collected[target].append(str(file_path))
return collected
def deduplicate(paths: list[str], threshold: int = 8) -> list[str]:
seen_hashes = []
unique = []
for path in paths:
image_hash = phash(path)
if not image_hash:
continue
is_duplicate = any(
sum(a != b for a, b in zip(image_hash, known_hash)) <= threshold
for known_hash in seen_hashes
)
if not is_duplicate:
seen_hashes.append(image_hash)
unique.append(path)
return unique
def reset_output_dir(output_dir: str) -> None:
root = Path(output_dir)
if not root.exists():
return
for split in SPLITS:
split_dir = root / split
if split_dir.exists():
shutil.rmtree(split_dir)
def split_and_copy(images: dict, output_dir: str) -> dict:
random.seed(42)
stats = {}
for cls, paths in images.items():
random.shuffle(paths)
total = len(paths)
n_train = int(total * SPLITS["train"])
n_val = int(total * SPLITS["val"])
split_paths = {
"train": paths[:n_train],
"val": paths[n_train:n_train + n_val],
"test": paths[n_train + n_val:],
}
for split, items in split_paths.items():
dest_dir = Path(output_dir) / split / cls
dest_dir.mkdir(parents=True, exist_ok=True)
for index, src in enumerate(items):
ext = Path(src).suffix.lower()
dest = dest_dir / f"{cls}_{split}_{index:05d}{ext}"
shutil.copy2(src, dest)
stats[cls] = {"total": total, **{split: len(items) for split, items in split_paths.items()}}
return stats
def main():
parser = argparse.ArgumentParser()
parser.add_argument(
"--trashnet_dir",
default="data/raw/trashnet",
help="Path to the unzipped TrashNet dataset",
)
parser.add_argument(
"--taco_dir",
default="data/raw/taco",
help="Path to the TACO image folders",
)
parser.add_argument(
"--realwaste_dir",
default="data/raw/realwaste",
help="Path to the organized RealWaste image folders",
)
parser.add_argument(
"--feedback_dir",
default="data/feedback_labeled",
help="Path to operator-reviewed feedback images organized by class",
)
parser.add_argument(
"--extra_dir",
default="data/local_boost",
help="Path to extra local training images organized by class",
)
parser.add_argument("--output_dir", default="data/processed")
args = parser.parse_args()
all_images = {c: [] for c in TARGET_CLASSES}
if os.path.isdir(args.trashnet_dir):
log.info("Collecting from TrashNet...")
trashnet_images = collect_images(args.trashnet_dir, TRASHNET_MAP)
for cls in TARGET_CLASSES:
all_images[cls].extend(trashnet_images[cls])
else:
log.warning(
"TrashNet dir not found: %s\nDownload from https://github.com/garythung/trashnet and unzip.",
args.trashnet_dir,
)
if os.path.isdir(args.taco_dir):
log.info("Collecting from TACO...")
taco_images = collect_images(args.taco_dir, TACO_MAP)
for cls in TARGET_CLASSES:
all_images[cls].extend(taco_images[cls])
else:
log.warning("TACO dir not found: %s. Skipping organic supplement.", args.taco_dir)
if os.path.isdir(args.realwaste_dir):
log.info("Collecting from RealWaste...")
realwaste_images = collect_images(args.realwaste_dir, REALWASTE_MAP)
for cls in TARGET_CLASSES:
all_images[cls].extend(realwaste_images[cls])
else:
log.warning("RealWaste dir not found: %s. Skipping RealWaste supplement.", args.realwaste_dir)
if os.path.isdir(args.feedback_dir):
log.info("Collecting from operator feedback...")
feedback_images = collect_images(args.feedback_dir, FEEDBACK_MAP)
for cls in TARGET_CLASSES:
all_images[cls].extend(feedback_images[cls])
else:
log.warning("Feedback dir not found: %s. Skipping operator feedback supplement.", args.feedback_dir)
if os.path.isdir(args.extra_dir):
log.info("Collecting from local boost dataset...")
local_images = collect_images(args.extra_dir, LOCAL_BOOST_MAP)
for cls in TARGET_CLASSES:
all_images[cls].extend(local_images[cls])
else:
log.warning("Local boost dir not found: %s. Skipping local boost supplement.", args.extra_dir)
log.info("Verifying images...")
for cls in TARGET_CLASSES:
before = len(all_images[cls])
all_images[cls] = [path for path in all_images[cls] if verify_image(path)]
removed = before - len(all_images[cls])
if removed:
log.warning(" %s: removed %s corrupted files", cls, removed)
log.info("Deduplicating...")
for cls in TARGET_CLASSES:
before = len(all_images[cls])
all_images[cls] = deduplicate(all_images[cls])
log.info(" %s: %s -> %s after dedup", cls, before, len(all_images[cls]))
log.info("Splitting and copying...")
reset_output_dir(args.output_dir)
stats = split_and_copy(all_images, args.output_dir)
print("\nDataset summary")
print(f"{'Class':<12} {'Total':>7} {'Train':>7} {'Val':>7} {'Test':>7}")
print("-" * 44)
for cls, summary in stats.items():
print(
f"{cls:<12} {summary['total']:>7} {summary['train']:>7} "
f"{summary['val']:>7} {summary['test']:>7}"
)
grand_total = sum(summary["total"] for summary in stats.values())
print(f"\nTotal images: {grand_total}")
print(f"Output dir : {os.path.abspath(args.output_dir)}")
if __name__ == "__main__":
main()
|