Request for original 192 kHz audio — Intel Robotic Welding Multimodal Dataset

#3
by Yumen38 - opened

Dear Dr. Lopez,

I am Yoonje Lee, a master's student at Hanyang University in South Korea, working on real-time multimodal weld defect detection for edge deployment.

I have been using the Intel Robotic Welding Multimodal Dataset alongside your paper (arXiv:2409.02290), and I wanted to thank you for releasing it — it is an exceptionally useful resource and has significantly started my work.

I have one question. The paper reports a 192 kHz sampling rate, while the released FLAC files are 16 kHz single-channel. Would it be possible to obtain the original 192 kHz recordings, or even a small subset?

Part of my work examines how much defect-relevant acoustic information lies above 8 kHz, which connects directly to your stated reason for selecting high-bandwidth microphones. A subset of a few dozen samples per defect category would be enough for the analysis, and I am glad to comply with any data sharing terms Intel requires.

If sharing is not possible, I completely understand — knowing that would also be helpful, as I could then plan to collect the high-frequency data myself.

Thank you for your time and for making this dataset available.

Best regards,

Yoonje Lee
M.S. Student, Department of Robotics
Hanyang University, South Korea
trueneat0223@hanyang.ac.kr

Sign up or log in to comment