Spaces:
Running
Running
File size: 8,128 Bytes
98847a8 9eea4a2 98847a8 9eea4a2 98847a8 9eea4a2 98847a8 013aff2 98847a8 503a577 98847a8 44072a9 9eea4a2 98847a8 503a577 98847a8 9eea4a2 98847a8 503a577 44072a9 9eea4a2 9bf569f 503a577 98847a8 503a577 98847a8 503a577 08dfd47 98847a8 44072a9 98847a8 503a577 44072a9 9eea4a2 9bf569f 503a577 98847a8 08dfd47 98847a8 08dfd47 98847a8 013aff2 98847a8 9eea4a2 98847a8 9eea4a2 98847a8 08dfd47 98847a8 08dfd47 503a577 08dfd47 ed37070 98847a8 08dfd47 013aff2 08dfd47 98847a8 9eea4a2 98847a8 9eea4a2 98847a8 9eea4a2 |
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 |
import ast
import json
import re
from pathlib import Path
import requests
from backend.config import get_example_config
def group_files_by_index(file_paths, data_type="audio"):
# Regular expression pattern to extract the key from each image path
if data_type == "audio":
pattern = r"audio_(\d+).(png|wav)"
elif data_type == "video":
pattern = r"video_(\d+).(png|mp4)"
else:
pattern = r"img_(\d+).png"
# Dictionary to store the grouped files
grouped_files = {}
# Iterate over each file path
for file_path in file_paths:
# Extract the key using the regular expression pattern
match = re.search(pattern, file_path)
if match:
key = int(match.group(1))
# Add the file path to the corresponding group in the dictionary
if key not in grouped_files:
grouped_files[key] = []
grouped_files[key].append(file_path)
# Sort the dictionary by keys
sorted_grouped_files = dict(sorted(grouped_files.items()))
return sorted_grouped_files
def build_description(
i, data_none, data_attack, quality_metrics=["psnr", "ssim", "lpips"]
):
# TODO: handle this at data generation
if isinstance(data_none["fake_det"], str):
data_none["fake_det"] = ast.literal_eval(data_none["fake_det"])
if isinstance(data_none["watermark_det"], str):
data_none["watermark_det"] = ast.literal_eval(data_none["watermark_det"])
if isinstance(data_attack["fake_det"], str):
data_attack["fake_det"] = ast.literal_eval(data_attack["fake_det"])
if isinstance(data_attack["watermark_det"], str):
data_attack["watermark_det"] = ast.literal_eval(data_attack["watermark_det"])
if i == 0:
fake_det = data_none["fake_det"]
return {"detected": fake_det}
elif i == 1:
# Fixed metrics
det = data_none["watermark_det"]
log10_p_value = float(data_none["log10_p_value"])
word_acc = data_attack["word_acc"]
bit_acc = data_none["bit_acc"]
# Dynamic metrics
metrics_output = {}
for metric in quality_metrics:
value = float(data_none[metric])
metrics_output[metric] = round(value, 3)
# Fixed metrics output
metrics_output.update(
{
"detected": det,
"log10_p_value": round(log10_p_value, 3),
"bit_acc": round(bit_acc, 3),
"word_acc": word_acc,
}
)
return metrics_output
elif i == 2:
fake_det = data_attack["fake_det"]
return {"detected": fake_det}
elif i == 3: # REVISIT THIS, it used to be == 3
det = data_attack["watermark_det"]
log10_p_value = float(data_attack["log10_p_value"])
word_acc = data_attack["word_acc"]
bit_acc = data_attack["bit_acc"]
return {
"detected": det,
"log10_p_value": round(log10_p_value, 3),
"bit_acc": round(bit_acc, 3),
"word_acc": word_acc,
}
def build_infos(abs_path: Path, datatype: str, dataset_name: str, db_key: str):
def generate_file_patterns(prefixes, extensions):
indices = [0, 1, 3, 4, 5]
return [
f"{prefix}_{index:05d}.{ext}"
for prefix in prefixes
for index in indices
for ext in extensions
]
if datatype == "audio":
quality_metrics = ["snr", "sisnr", "stoi", "pesq"]
extensions = ["wav"]
datatype_abbr = "audio"
eval_results_path = abs_path + f"{dataset_name}_1k/examples_eval_results.json"
elif datatype == "image":
quality_metrics = ["psnr", "ssim", "lpips"]
extensions = ["png"]
datatype_abbr = "img"
eval_results_path = abs_path + f"{dataset_name}_1k/examples_eval_results.json"
elif datatype == "video":
quality_metrics = ["psnr", "ssim", "lpips", "msssim", "vmaf"]
extensions = ["mp4"]
datatype_abbr = "video"
eval_results_path = abs_path + f"{dataset_name}/examples_eval_results.json"
# Determine if eval_results_path is a URL or local file
if eval_results_path.startswith("http://") or eval_results_path.startswith(
"https://"
):
response = requests.get(eval_results_path)
if response.status_code == 200:
results_data = response.json()
else:
return {}
else:
try:
with open(eval_results_path, "r") as f:
results_data = json.load(f)
except Exception as e:
print(f"Failed to load local file: {e}")
return {}
dataset = results_data["eval"][db_key]
prefixes = [
f"attacked_{datatype_abbr}",
f"attacked_wmd_{datatype_abbr}",
f"{datatype_abbr}",
f"wmd_{datatype_abbr}",
]
file_patterns = generate_file_patterns(prefixes, extensions)
infos = {}
for model_name in dataset.keys():
model_infos = {}
default_attack_name = "none"
if datatype == "audio":
default_attack_name = "identity"
elif datatype == "video":
default_attack_name = "Identity"
identity_attack_rows = dataset[model_name][default_attack_name]["default"]
for attack_name, attack_variants_data in dataset[model_name].items():
for attack_variant, attack_rows in attack_variants_data.items():
if attack_variant == "default":
attack = attack_name
else:
attack = f"{attack_name}_{attack_variant}"
if len(attack_rows) == 0:
model_infos[attack] = []
continue
if datatype == "video":
file_paths = [
f"{abs_path}{dataset_name}/examples/{datatype}/{model_name}/{attack}/{pattern}"
for pattern in file_patterns
]
else:
file_paths = [
f"{abs_path}{dataset_name}_1k/examples/{datatype}/{model_name}/{attack}/{pattern}"
for pattern in file_patterns
]
all_files = []
for i, files in group_files_by_index(
file_paths,
data_type=datatype,
).items():
data_none = [e for e in identity_attack_rows if e["idx"] == i][0]
data_attack = [e for e in attack_rows if e["idx"] == i][0]
files = sorted(
[(f, Path(f).stem) for f in files], key=lambda x: x[1]
)
files = files[2:] + files[:2]
new_files = []
for variant_i, (file, name) in enumerate(files):
file_info = {
"name": name,
"metadata": build_description(
variant_i, data_none, data_attack, quality_metrics
),
}
if datatype == "audio":
file_info["image_url"] = file.replace(".wav", ".png")
file_info["audio_url"] = file
elif datatype == "video":
# file_info["image_url"] = file.replace(".mp4", ".png")
file_info["video_url"] = file
else:
file_info["image_url"] = file
new_files.append(file_info)
all_files.extend(new_files)
model_infos[attack] = all_files
infos[model_name] = model_infos
return infos
def get_examples_tab(datatype: str):
config = get_example_config(datatype)
infos = build_infos(
config["path"],
datatype=datatype,
dataset_name=config["dataset_name"],
db_key=config["db_key"],
)
return infos
|