audio-datasets/get_diff.py at main · GLJS/audio-datasets · GitHub

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
import pickle
import os
import pandas as pd
import numpy as np
import torch
from torch.utils.data import Dataset, DataLoader
import json
from tqdm import tqdm
import itertools

# Load allowed datasets from unique_datasets.json
with open('visualization/unique_datasets.json', 'r') as f:
    known_datasets = json.load(f)

class EmbeddingDataset(Dataset):
    def __init__(self, embeddings):
        self.embeddings = torch.tensor(embeddings)

    def __len__(self):
        return len(self.embeddings)

    def __getitem__(self, idx):
        return self.embeddings[idx]

def load_embeddings(embeddings_dir):
    """Load audio embeddings"""
    embeddings_path = os.path.join(embeddings_dir, 'audio')
    dataset_embeddings = {}

    for dataset in os.listdir(embeddings_path):
        if dataset not in known_datasets:
            continue

        # Load name mapping if it exists
        name_map = {}
        name_map_path = f"name_maps/{dataset}_name_map.json"
        if os.path.exists(name_map_path):
            with open(name_map_path, 'r') as f:
                loaded_map = json.load(f)
                # Flatten the name map by combining train/valid/test dicts
                name_map = {}
                for split in loaded_map.keys():
                    name_map.update(loaded_map[split])

        dataset_path = os.path.join(embeddings_path, dataset)
        embeddings = []
        filenames = []

        for split_file in os.listdir(dataset_path):
            with open(os.path.join(dataset_path, split_file), 'rb') as f:
                data = pickle.load(f)
                files, embeds = zip(*data)
                embeddings.extend(embeds)
                filenames.extend(files)

        dataset_embeddings[dataset] = {
            'embeddings': np.array(embeddings),
            'filenames': filenames,
            'name_map': name_map
        }

    return dataset_embeddings

def compute_similarities(embeddings_dict, batch_size=32768):
    """Compute pairwise similarities between datasets"""
    os.makedirs("splits", exist_ok=True)
    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
    pbar = tqdm(list(itertools.combinations(embeddings_dict.keys(), 2)),
                desc="Computing audio similarities")
    for d1, d2 in pbar:
        pbar.set_description(f"Computing audio similarities between {d1} and {d2}")
        # Create dataloaders
        dataset1 = EmbeddingDataset(embeddings_dict[d1]['embeddings'])
        dataset2 = EmbeddingDataset(embeddings_dict[d2]['embeddings'])
        loader1 = DataLoader(dataset1, batch_size=batch_size, num_workers=8)
        loader2 = DataLoader(dataset2, batch_size=batch_size, num_workers=8)

        similar_pairs_d1 = []
        similar_pairs_d2 = []

        # Compute similarities in batches
        for i, batch1 in enumerate(tqdm(loader1, leave=False)):
            batch1 = batch1.to(device)
            start_idx1 = i * batch_size

            for j, batch2 in enumerate(loader2):
                batch2 = batch2.to(device)
                start_idx2 = j * batch_size

                batch1_norm = batch1 / batch1.norm(dim=1, keepdim=True)
                batch2_norm = batch2 / batch2.norm(dim=1, keepdim=True)

                # Compute cosine similarities
                sims = torch.mm(batch1_norm, batch2_norm.t())

                # Find pairs with >99% similarity
                similar_idx = torch.where(sims > 0.99)
                batch1_idx = similar_idx[0] + start_idx1
                batch2_idx = similar_idx[1] + start_idx2

                # Get filenames for similar pairs
                for idx1, idx2 in zip(batch1_idx.cpu(), batch2_idx.cpu()):
                    similar_pairs_d1.append(str(embeddings_dict[d1]['filenames'][idx1]))
                    similar_pairs_d2.append(str(embeddings_dict[d2]['filenames'][idx2]))

        if len(similar_pairs_d1) > 0:
            # Apply name mapping
            similar_pairs_d1_mapped = [str(embeddings_dict[d1]['name_map'].get(f, f)) for f in similar_pairs_d1]
            similar_pairs_d2_mapped = [str(embeddings_dict[d2]['name_map'].get(f, f)) for f in similar_pairs_d2]

            similar_pairs_d1_short = ["/".join(f.split('/')[-2:]) if len(f.split("/")) > 2 else f for f in similar_pairs_d1_mapped]
            similar_pairs_d2_short = ["/".join(f.split('/')[-2:]) if len(f.split("/")) > 2 else f for f in similar_pairs_d2_mapped]

            similar_pairs_d1_short = list(set(similar_pairs_d1_short))
            similar_pairs_d2_short = list(set(similar_pairs_d2_short))

            # Save similar pairs to CSV files
            pd.DataFrame({'file_name': similar_pairs_d1_short}).to_csv(
                f'splits/{d1.lower()}_in_{d2.lower()}.csv', index=False)
            pd.DataFrame({'file_name': similar_pairs_d2_short}).to_csv(
                f'splits/{d2.lower()}_in_{d1.lower()}.csv', index=False)

def main():
    embeddings_dir = "/gpfs/work4/0/einf6190/embeddings"
    embeddings_dict = load_embeddings(embeddings_dir)
    compute_similarities(embeddings_dict)

if __name__ == "__main__":
    main()