search

mdda / deep-learning-workshop

Deep Learning Workshop : Including a VirtualBox VM with pre-configured Jupyter, Tensorflow, PyTorch, models and data
MIT License
470 stars 121 forks source link

num.pkl dataset missing #3

Open dorienh opened 7 years ago

dorienh commented 7 years ago

Hello,

Thank you for this code. I enjoyed your talk at Google Singapore a while back.

I was trying to run SpeechRecognition_Learn.ipynb but I could not find the num.pkl or test dataset. I was just trying to understand the data format it is loaded in.

The train_data also doesn't seem to be defined.

dorienh commented 7 years ago

I actually see that there is a script to generate the data. Nevertheless, it would be great to see the spoken letters pickle. The generation script seems to be a bit different with words and sentences. Thanks already for the insightful tutorials.

mdda commented 7 years ago

I'll add some links to a pre-done training/test set in the folder's README.md. But I don't really want to embed the downloader in the notebooks, since my server isn't really intended to be a data repository...

dorienh commented 7 years ago

I understand. I was mostly interesting in understanding the structure of the pickle file.

On 25 July 2017 at 11:52, Martin Andrews notifications@github.com wrote:

I'll add some links to a pre-done training/test set. But I don't really want to embed the downloader in the notebooks, since my server isn't really intended to be a data repository...

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/mdda/deep-learning-workshop/issues/3#issuecomment-317621324, or mute the thread https://github.com/notifications/unsubscribe-auth/AEhJe0MrZoIJpXE1gklT8Za23CdXSmeQks5sRWZlgaJpZM4Ogrfv .

-- Dorien Herremans, PhD Marie-Curie Fellow http://dorienherremans.com

Queen Mary University of London School of Electronic Engineering and Computer Science C4DM - Centre for Digital Music, London

Workshop on Deep Learning and Music http://dorienherremans.com/dlm2017, May'17 Anchorage, Alaska.

dorienh commented 7 years ago

After studying the code for a bit, it seems to be a bit different then I remember from the meetup. I remember something about an SVM to correct errors afterwards, but I didn't see that. I'm also slightly confused still by the dataset and how the images are created from your voice saying the letters. Are they the stamps? Or do you use the spectograms.

Sorry for my many questions and thanks in advance for the clarification.

mdda commented 7 years ago

Does the explanation in the folder's README.md make sense to you?