Back to datasets
Dataset assetOpen Source CommunityAudio ProcessingSound Source Localization

FidelOdok/SOFA_DOA_10_deg_meta

This dataset comprises audio recordings together with corresponding labels, direction vectors and angle measurements. Labels range from 0 to 96. The training split contains 22,500 samples, totalling 21.5 GB; the download size is approximately 21.5 GB.

Source
hugging_face
Created
Nov 28, 2025
Updated
Jun 16, 2023
Signals
126 views
Availability
Linked source ready
Overview

Dataset description and usage context

Dataset Overview

Feature Information

  • audio: audio data, type audio.
  • label: class label, type class_label, with categories:
    • 0: 0
    • 1: 101
    • 2: 106
    • 3: 112
    • 4: 117
    • 5: 122
    • 6: 129
    • 7: 134
    • 8: 137
    • 9: 139
    • 10: 151
    • 11: 156
    • 12: 166
    • 13: 169
    • 14: 171
    • 15: 172
    • 16: 18
    • 17: 182
    • 18: 187
    • 19: 189
    • 20: 190
    • 21: 192
    • 22: 200
    • 23: 205
    • 24: 207
    • 25: 209
    • 26: 211
    • 27: 218
    • 28: 219
    • 29: 221
    • 30: 224
    • 31: 226
    • 32: 227
    • 33: 229
    • 34: 237
    • 35: 239
    • 36: 242
    • 37: 244
    • 38: 257
    • 39: 26
    • 40: 260
    • 41: 262
    • 42: 265
    • 43: 278
    • 44: 281
    • 45: 3
    • 46: 312
    • 47: 317
    • 48: 328
    • 49: 343
    • 50: 351
    • 51: 354
    • 52: 356
    • 53: 358
    • 54: 359
    • 55: 368
    • 56: 369
    • 57: 371
    • 58: 372
    • 59: 373
    • 60: 378
    • 61: 380
    • 62: 383
    • 63: 385
    • 64: 386
    • 65: 391
    • 66: 394
    • 67: 397
    • 68: 4
    • 69: 422
    • 70: 423
    • 71: 424
    • 72: 426
    • 73: 427
    • 74: 428
    • 75: 46
    • 76: 49
    • 77: 5
    • 78: 50
    • 79: 58
    • 80: 6
    • 81: 66
    • 82: 67
    • 83: 69
    • 84: 7
    • 85: 71
    • 86: 73
    • 87: 82
    • 88: 84
    • 89: 86
    • 90: 87
    • 91: 89
    • 92: 96
  • dirx: direction x, type float64.
  • diry: direction y, type float64.
  • dirz: direction z, type float64.
  • phi: angle phi, type float64.
  • theta: angle theta, type float64.

Data Splits

  • train: training split with 22,500 samples, total bytes 21,492,687,405.0.

Dataset Size

  • Download size: 21,493,917,608 bytes
  • Dataset size: 21,492,687,405.0 bytes
Need downstream help?

Pair the dataset with AI analysis and content workflows.

Once the source passes your review, move straight into summarization, transformation, report drafting, or presentation generation with the JuheAI toolchain.

Explore AI studio