omnisealbench / backend /config.py
Mark Duppenthaler
Add dataset selector
54be5f9
raw
history blame
4.17 kB
def get_dataset_config(dataset_name):
if dataset_name == "voxpopuli_1k_audio_benchmark":
return {
"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",
},
}
elif dataset_name == "ravdess_1k_audio_benchmark":
return {
"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",
},
}
elif dataset_name == "val2014_1k_image_benchmark":
return {
"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",
},
}
elif dataset_name == "sav_val_full_video_benchmark":
return {
"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",
},
}
else:
raise ValueError(f"Unknown dataset: {dataset_name}")