ABS_DATASET_DOMAIN = "https://dl.fbaipublicfiles.com" ABS_DATASET_PATH = f"{ABS_DATASET_DOMAIN}/omnisealbench/" def get_dataset_config(dataset_name): if dataset_name == "voxpopuli_1k/audio": 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", }, "attacks_with_variations": [ "random_noise", "lowpass_filter", "highpass_filter", "boost_audio", "duck_audio", "shush", ], } elif dataset_name == "ravdess_1k/audio": 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", }, "attacks_with_variations": [ "random_noise", "lowpass_filter", "highpass_filter", "boost_audio", "duck_audio", "shush", ], } elif dataset_name == "val2014_1k/image": 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", }, "attacks_with_variations": [ "center_crop", "jpeg", "brightness", "contrast", "saturation", "sharpness", "resize", "perspective", "median_filter", "hue", "gaussian_blur", ], } elif dataset_name == "sav_val_full/video": 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", }, "attacks_with_variations": [ "Rotate", "Resize", "Crop", "Brightness", "Contrast", "Saturation", "H264", "H264rgb", "H265", ], } else: raise ValueError(f"Unknown dataset: {dataset_name}")