cmudeeplearning11785 / spring2019_tutorials Goto Github PK

I'm interested in HW1 part 2 (Input Frame Mel features -> Predict Phoneme state), and I want to test my model on my audio voice data.

According to kaggle S2020 course

To convert the speech to a mel-spectrogram, it is segmented into little "frames", each 25ms wide, where the "stride" between adjacent frames is 10ms. Thus we get 100 such frames per second of speech.

Thus, for a T-second recording, the entire spectrogram is a 100T x 40 matrix, comprising 100T 40- dimensional vectors (at 100 vectors (frames) per second).

But the dataset provided on this homework is already converted to mel-spectrogram without further explain on the process and other config of mel (Apply preprocessing, resample, normalize amplitude, frequency range).

Can you please give the source to generate your dataset ? Thank you

Thanks for sharing these series of notebook with us.
I just started and got stuck by

"from dataloader import get_data"

I wonder if this is an external file that is missing from the folder or this refers to the dataloader package ? if the later, the get_data does not exist.

---------------------------------------------------------------------------
ImportError                               Traceback (most recent call last)
<ipython-input-6-80d340b35860> in <module>()
      1 import pickle,csv
----> 2 from dataloader import get_data

/Users/opt/anaconda3/envs/CMU/lib/python3.7/site-packages/dataloader/__init__.py in <module>()
----> 1 from dataloader import read_data_sets

ImportError: cannot import name 'read_data_sets' from 'dataloader' (/Users/opt/anaconda3/envs/CMU/lib/python3.7/site-packages/dataloader/__init__.py)

I also looked into the previous years but seems nobody really had a similar issue. I appreciate your comments.

Could you please open source the assignment answers, I have done assignments but I do not know whether it is right. I would really appreciate it if you could help me.