File size: 8,279 Bytes
0e5ae7a 4a0bcb8 0e5ae7a 4a0bcb8 0e5ae7a |
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 |
import os
import click
import json
from PIL import Image
from tqdm import tqdm
def resize_image(image: Image, size: int):
# resize the image so that the shorter side is `size` pixels
w, h = image.size
if w < h:
new_w = size
new_h = int(size * h / w)
else:
new_w = int(size * w / h)
new_h = size
return image.resize((new_w, new_h))
SUPPORTED_WEATHERS = ["clear", "rainy", "foggy", "snowy"]
SUPPORTED_TIMESOFDAY = ["daytime", "night"]
WEATHER_MAPPING = {
"clear": "clear",
"rainy": "rain",
"foggy": "fog",
"snowy": "snow",
}
@click.command()
@click.option("--bdd100k_dir", type=str, default="datasets/bdd100k")
@click.option("--acdc_dir", type=str, default="datasets/acdc")
@click.option("--output_dir", type=str, default="datasets/swim_data")
def build_swim_dataset(bdd100k_dir: str, acdc_dir: str, output_dir: str):
# build the dataset with format
# swim_data
# ├── train
# │ ├── images
# │ │ ├── 000000.jpg
# │ │ ├── ...
# │ ├── labels.json
# ├── val
# │ ├── images
# │ │ ├── 000000.jpg
# │ │ ├── ...
# │ ├── labels.json
#
# labels.json
# [
# {
# "name": "000000.jpg",
# "weather": "clear"|"rain"|"fog"|"snow",
# "timeofday": "daytime"|"night",
# "source": "bdd100k"|"acdc",
# }
# ]
#
# note:
# - all images are resized so that the shorter side is 512 pixels
# - train: bdd100k train + acdc train + acdc val
# - val: bdd100k val + acdc test
#
# bdd100k format:
# bdd100k
# ├── images
# │ ├── 100k
# │ │ ├── train
# │ │ │ ├── 000000.jpg
# │ │ │ ├── ...
# │ │ ├── val
# │ │ │ ├── 000000.jpg
# │ │ │ ├── ...
# ├── labels
# │ ├── det_20
# │ │ ├── det_train.json
# │ │ ├── det_val.json
#
# acdc format:
# acdc
# ├── fog
# │ ├── test
# │ │ ├── <subfolder>
# │ │ │ ├── 000000.jpg
# │ │ │ ├── ...
# │ ├── test_ref (clear images)
# │ │ ├── <subfolder>
# │ │ │ ├── 000000.jpg
# │ │ │ ├── ...
# | ├── train
# │ │ ├── <subfolder>
# │ │ │ ├── 000000.jpg
# │ │ │ ├── ...
# | ├── train_ref (clear images)
# │ │ ├── <subfolder>
# │ │ │ ├── 000000.jpg
# │ │ │ ├── ...
# | ├── val
# │ │ ├── <subfolder>
# │ │ │ ├── 000000.jpg
# │ │ │ ├── ...
# | ├── val_ref (clear images)
# │ │ ├── <subfolder>
# ├── rain
# ├── snow
# ├── night
os.makedirs(output_dir, exist_ok=True)
os.makedirs(os.path.join(output_dir, "train", "images"), exist_ok=True)
os.makedirs(os.path.join(output_dir, "val", "images"), exist_ok=True)
count = 0
# build train dataset
labels = []
# bdd100k train
with open(os.path.join(bdd100k_dir, "labels", "det_20", "det_train.json")) as f:
bdd100k_train_labels = json.load(f)
for label in tqdm(bdd100k_train_labels, desc="bdd100k train"):
if label["attributes"]["weather"] not in SUPPORTED_WEATHERS:
continue
if label["attributes"]["timeofday"] not in SUPPORTED_TIMESOFDAY:
continue
new_name = f"{count:06d}.jpg"
count += 1
image = Image.open(
os.path.join(bdd100k_dir, "images", "100k", "train", label["name"])
)
image = resize_image(image, 512)
image.save(os.path.join(output_dir, "train", "images", new_name))
labels.append(
{
"name": new_name,
"weather": WEATHER_MAPPING[label["attributes"]["weather"]],
"timeofday": label["attributes"]["timeofday"],
"source": "bdd100k",
}
)
# acdc train + val
for weather in ["fog", "rain", "snow", "night"]:
for split in ["train", "val", "train_ref", "val_ref"]:
for subfolder in os.listdir(os.path.join(acdc_dir, weather, split)):
for filename in os.listdir(
os.path.join(acdc_dir, weather, split, subfolder)
):
new_name = f"{count:06d}.jpg"
count += 1
image = Image.open(
os.path.join(acdc_dir, weather, split, subfolder, filename)
)
image = resize_image(image, 512)
image.save(os.path.join(output_dir, "train", "images", new_name))
if split.endswith("ref"):
timeofday_ = "daytime"
weather_ = "clear"
else:
timeofday_ = "night" if weather == "night" else "daytime"
weather_ = "clear" if weather == "night" else weather
labels.append(
{
"name": new_name,
"weather": weather_,
"timeofday": timeofday_,
"source": "acdc",
}
)
with open(os.path.join(output_dir, "train", "labels.json"), "w") as f:
json.dump(labels, f, indent=4)
# build val dataset
labels = []
# bdd100k val
with open(os.path.join(bdd100k_dir, "labels", "det_20", "det_val.json")) as f:
bdd100k_val_labels = json.load(f)
for label in tqdm(bdd100k_val_labels, desc="bdd100k val"):
if label["attributes"]["weather"] not in SUPPORTED_WEATHERS:
continue
if label["attributes"]["timeofday"] not in SUPPORTED_TIMESOFDAY:
continue
new_name = f"{count:06d}.jpg"
count += 1
image = Image.open(
os.path.join(bdd100k_dir, "images", "100k", "val", label["name"])
)
image = resize_image(image, 512)
image.save(os.path.join(output_dir, "val", "images", new_name))
labels.append(
{
"name": new_name,
"weather": WEATHER_MAPPING[label["attributes"]["weather"]],
"timeofday": label["attributes"]["timeofday"],
"source": "bdd100k",
}
)
# acdc test
for weather in ["fog", "rain", "snow", "night"]:
for split in ["test", "test_ref"]:
for subfolder in os.listdir(os.path.join(acdc_dir, weather, split)):
for filename in os.listdir(
os.path.join(acdc_dir, weather, split, subfolder)
):
new_name = f"{count:06d}.jpg"
count += 1
image = Image.open(
os.path.join(acdc_dir, weather, split, subfolder, filename)
)
image = resize_image(image, 512)
image.save(os.path.join(output_dir, "val", "images", new_name))
if split.endswith("ref"):
timeofday_ = "daytime"
weather_ = "clear"
else:
timeofday_ = "night" if weather == "night" else "daytime"
weather_ = "clear" if weather == "night" else weather
labels.append(
{
"name": new_name,
"weather": weather_,
"timeofday": timeofday_,
"source": "acdc",
}
)
with open(os.path.join(output_dir, "val", "labels.json"), "w") as f:
json.dump(labels, f, indent=4)
if __name__ == "__main__":
build_swim_dataset()
|