Spaces:
Running
Running
File size: 8,322 Bytes
9eea4a2 6a879a9 cc391f1 6a879a9 b087e88 9eea4a2 b087e88 6a879a9 b5bd664 6a879a9 b5bd664 6a879a9 b5bd664 cca4f79 6a879a9 b5bd664 6a879a9 cc391f1 6a879a9 cca4f79 6a879a9 b5bd664 6a879a9 b5bd664 b087e88 cca4f79 cc391f1 cca4f79 9eea4a2 cca4f79 6a879a9 cca4f79 6a879a9 cca4f79 cc391f1 cca4f79 6a879a9 9eea4a2 cc391f1 cca4f79 54be5f9 b5bd664 6a879a9 54be5f9 6a879a9 cc391f1 6a879a9 cc391f1 6a879a9 54be5f9 |
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 |
# Change these values to match your dataset structure if loading locally or from a different source.
# IMPORTANT: When running from docker more setup is required (e.g. on Huggingface)
import os
from collections import defaultdict
from pathlib import Path
ABS_DATASET_DOMAIN = "https://dl.fbaipublicfiles.com"
# Sample dataset domain and path for local loading
# Some more configuration may be required to load examples from local files.
# ABS_DATASET_DOMAIN = "./data"
ABS_DATASET_PATH = f"{ABS_DATASET_DOMAIN}/omnisealbench/"
MODALITY_CONFIG_CONSTANTS = {
"audio": {
"first_cols": [
"snr",
"sisnr",
"stoi",
"pesq",
],
"attack_scores": [
"bit_acc",
"log10_p_value",
"TPR",
"FPR",
],
"categories": {
"speed": "Time",
"updownresample": "Time",
"echo": "Time",
"random_noise": "Amplitude",
"lowpass_filter": "Amplitude",
"highpass_filter": "Amplitude",
"bandpass_filter": "Amplitude",
"smooth": "Amplitude",
"boost_audio": "Amplitude",
"duck_audio": "Amplitude",
"shush": "Amplitude",
"pink_noise": "Amplitude",
"aac_compression": "Compression",
"mp3_compression": "Compression",
},
"attacks_with_variations": [
"random_noise",
"lowpass_filter",
"highpass_filter",
"boost_audio",
"duck_audio",
"shush",
],
},
"image": {
"first_cols": ["psnr", "ssim", "lpips", "decoder_time"],
"attack_scores": ["bit_acc", "log10_p_value", "TPR", "FPR"],
"categories": {
"proportion": "Geometric",
"collage": "Inpainting",
"center_crop": "Geometric",
"rotate": "Geometric",
"jpeg": "Compression",
"brightness": "Visual",
"contrast": "Visual",
"saturation": "Visual",
"sharpness": "Visual",
"resize": "Geometric",
"overlay_text": "Inpainting",
"hflip": "Geometric",
"perspective": "Geometric",
"median_filter": "Visual",
"hue": "Visual",
"gaussian_blur": "Visual",
"comb": "Mixed",
"avg": "Averages",
"none": "Baseline",
},
"attacks_with_variations": [
"center_crop",
"jpeg",
"brightness",
"contrast",
"saturation",
"sharpness",
"resize",
"perspective",
"median_filter",
"hue",
"gaussian_blur",
],
},
"video": {
"first_cols": ["psnr", "ssim", "msssim", "lpips", "vmaf", "decoder_time"],
"attack_scores": ["bit_acc", "log10_p_value", "TPR", "FPR"],
"categories": {
"HorizontalFlip": "Geometric",
"Rotate": "Geometric",
"Resize": "Geometric",
"Crop": "Geometric",
"Perspective": "Geometric",
"Brightness": "Visual",
"Contrast": "Visual",
"Saturation": "Visual",
"Grayscale": "Visual",
"Hue": "Visual",
"JPEG": "Compression",
"GaussianBlur": "Visual",
"MedianFilter": "Visual",
"H264": "Compression",
"H264rgb": "Compression",
"H265": "Compression",
"VP9": "Compression",
"H264_Crop_Brightness0": "Mixed",
"H264_Crop_Brightness1": "Mixed",
"H264_Crop_Brightness2": "Mixed",
"H264_Crop_Brightness3": "Mixed",
},
"attacks_with_variations": [
"Rotate",
"Resize",
"Crop",
"Brightness",
"Contrast",
"Saturation",
"H264",
"H264rgb",
"H265",
],
},
}
DATASET_CONFIGS = {
"voxpopuli_1k/audio": {"type": "audio", "path": ABS_DATASET_PATH},
"ravdess_1k/audio": {"type": "audio", "path": ABS_DATASET_PATH},
"val2014_1k_v2/image": {"type": "image", "path": ABS_DATASET_PATH},
"sa_1b_val_1k/image": {"type": "image", "path": ABS_DATASET_PATH},
"sav_val_full_v2/video": {"type": "video", "path": ABS_DATASET_PATH},
}
def get_user_dataset():
datasets = defaultdict(list)
default_local_data_dir = str(Path(__file__).parent.joinpath("data"))
user_data_dir = os.getenv("OMNISEAL_LEADERBOARD_DATA", default_local_data_dir)
if user_data_dir:
for user_data in os.listdir(user_data_dir):
if not os.path.isdir(os.path.join(user_data_dir, user_data)):
continue
user_dtype = os.listdir(os.path.join(user_data_dir, user_data, "examples"))[
0
]
datasets[user_dtype].append(user_data + "/" + user_dtype)
return datasets
def get_datasets():
grouped = {"audio": [], "image": [], "video": []}
for name, cfg in DATASET_CONFIGS.items():
dtype = cfg.get("type")
if dtype in grouped:
grouped[dtype].append(name)
# Add user datasets
user_datasets = get_user_dataset()
for dtype, user_names in user_datasets.items():
if dtype in grouped:
_names = [name for name in user_names if name not in grouped[dtype]]
grouped[dtype].extend(_names)
return grouped
def get_example_config(type, dataset_name):
"""Get example configuration for a specific dataset."""
if not dataset_name:
raise ValueError("Dataset name is required")
# Check if it's a valid dataset for this type
all_datasets = get_datasets()
if dataset_name not in all_datasets.get(type, []):
raise ValueError(f"Unknown dataset {dataset_name} for type {type}")
# Extract the dataset name without the type suffix
dataset_base_name = dataset_name.split("/")[0]
# Check if it's a user dataset
user_datasets = get_user_dataset()
user_data_dir = os.getenv("OMNISEAL_LEADERBOARD_DATA", "./data")
if dataset_name in user_datasets.get(type, []):
# It's a user dataset
examples_config = {
"dataset_name": dataset_base_name,
"path": user_data_dir + "/",
"db_key": dataset_base_name,
}
else:
# It's a predefined dataset from DATASET_CONFIGS
if dataset_name in DATASET_CONFIGS:
dataset_config = DATASET_CONFIGS[dataset_name]
examples_config = {
"dataset_name": dataset_base_name,
"path": dataset_config["path"],
"db_key": _get_db_key_for_dataset(dataset_base_name),
}
else:
raise ValueError(f"Dataset {dataset_name} not found in configurations")
return examples_config
def _get_db_key_for_dataset(dataset_base_name):
"""Helper function to determine the database key for a dataset"""
# Map of dataset names to their db keys
db_key_mapping = {
"voxpopuli_1k": "voxpopuli",
"val2014_1k_v2": "local_val2014",
"sa_1b_val_1k": "local_valid",
"sav_val_full_v2": "sa-v_sav_val_videos",
"ravdess_1k": "ravdess", # Add mapping for ravdess dataset
}
return db_key_mapping.get(dataset_base_name, dataset_base_name)
def get_dataset_config(dataset_name):
if dataset_name in DATASET_CONFIGS:
cfg = DATASET_CONFIGS[dataset_name]
extra_cfg = MODALITY_CONFIG_CONSTANTS.get(cfg["type"], {})
cfg.update(extra_cfg)
return cfg
else:
modality = dataset_name.split("/")[-1]
user_dataset = get_user_dataset()
default_local_data_dir = str(Path(__file__).parent.joinpath("data"))
user_data_dir = os.getenv("OMNISEAL_LEADERBOARD_DATA", default_local_data_dir)
if dataset_name in user_dataset.get(modality, []):
cfg = {
"type": modality,
"path": user_data_dir,
}
extra_cfg = MODALITY_CONFIG_CONSTANTS.get(cfg["type"], {})
cfg.update(extra_cfg)
return cfg
raise ValueError(f"Unknown dataset: {dataset_name}")
|