Dataset Viewer
Auto-converted to Parquet Duplicate
audio
audioduration (s)
20
40
label
stringlengths
6
15
user_hash
stringclasses
1 value
phase
int64
1
12
db_rms
float64
-71.2
-16.5
background
846c53d25527881d
1
-71.2
idle_breathing
846c53d25527881d
2
-56.3
antenna_slow
846c53d25527881d
3
-31
antenna_medium
846c53d25527881d
4
-22.5
antenna_fast
846c53d25527881d
5
-16.5
head_x
846c53d25527881d
6
-54.3
head_y
846c53d25527881d
7
-50.5
head_z
846c53d25527881d
8
-59.1
head_yaw
846c53d25527881d
9
-50.9
head_pitch
846c53d25527881d
10
-50.6
dance_smooth
846c53d25527881d
11
-28.1
dance_collision
846c53d25527881d
12
-24.3

Reachy RENS - Motor Noise Dataset

Crowdsourced recordings of Reachy Mini robot motor noise for training ego-noise suppression models.

Purpose

This dataset enables training of real-time motor noise cancellation models (DTLN) that can:

  • Remove motor whine from live audio streams
  • Improve audio analysis (beat detection, speech recognition)
  • Create cleaner livestream audio for robotics applications

Structure

data/
├── {user_hash_1}/
│   ├── background.wav    # 15s room tone (silence)
│   ├── idle.wav          # 15s idle/breathing motor noise
│   ├── active.wav        # 30s active dance motor noise
│   └── metadata.json     # Recording metadata
├── {user_hash_2}/
│   └── ...

Recording Specs

  • Sample Rate: 16kHz
  • Channels: Mono
  • Format: WAV (16-bit PCM)
  • Total Duration: ~60s per contributor

Contributing

Use the RENS data collector script:

pip install huggingface_hub sounddevice soundfile requests
huggingface-cli login
python rens_data_collector.py

Privacy

  • Contributors are identified by anonymized hash only
  • No personal information is collected
  • Audio is reviewed locally before upload consent

License

CC-BY-4.0 - Free to use with attribution

Project

Part of the RENS (Reachy Ego-Noise Suppression) project. https://github.com/TwinPeaksTownie/reachy-mini

Downloads last month
10

Space using TwinPeaksTownie/Reachy_RENS 1