New Kaggle Comp: Tensorflow Speech Recognition Challenge

Part 1 (2018)
1

Hey guys,

Check out this new Kaggle competition that just launched today! I believe we should be able to use convnets to classify the spoken word from the speech sound wave images (aka spectograms)… should be an interesting challenge!

Just discovered they are also giving away $500 in GCP (Google Cloud) credits for working on this competition. Details below :slight_smile:

https://www.kaggle.com/speech-commands-gcp-credits

Rules:

  • First 500 participants/teams to enter are eligible.
  • Entries will only be accepted until December 5th, 2017 11:59PM UTC.
  • Coupons will be issued within a week thereafter.
  • You must redeem and use the credit by the end of the competition.
11 Likes
2

Looks like you can convert the audio wav files to spectogram using scipy
https://docs.scipy.org/doc/scipy-0.19.0/reference/generated/scipy.signal.spectrogram.html

5 Likes
3

Would anyone like to form a team to tackle this challenge?

4

i’m interested,but i’m a newbie in the field of deep learning will any of the experienced practitioner of deep learning accept me as their team member .so,that i can grow,learn and achieve new heights with them in the domain of deep learning through this competition.

1 Like
5

interested, just ordered Raspberry Pi from Amazon, and this competition offers $500 GCP credits.

1 Like
6

Also interested.

1 Like
7

@hongta which model of pi

8

Pi3, They have provided a link to amazon for buying this

9

i already have pi3 model b ,

10

I know @Ekami has some experience working in Tensorflow on RPI3. Tuatini did you write one of your nice guides on that? :slight_smile:

2 Likes
11

@naveenmanwani or you can use emulater

2 Likes
12

I did, here is the guide for those who are interested :slight_smile: ( I should probably update it to TF 1.4 someday)

5 Likes
13

I’m interested.
I have some experience with rpi3.
Im still learning basics of Dl an tf library though :sweat_smile:

1 Like
14

that’s a great guide .one question it’s being said that you should not remove 2.7 python from pi ,because of some dependencies issue.but i want to remove 2.7 completely so that i should only work with python 3 with tensorflow on pi.is there any solution towards it

15

The 2 easiest ways I know to do this is just to install anaconda for python 3 (It will do the necessary replacement of your path from python 2 to python 3) or you could replace /usr/bin/python (which is a symlink pointing to python2.7) from your machine to python3.
Example:

~ ➜ ls -l /usr/bin/python
lrwxrwxrwx 1 root root 9 janv.  7  2017 /usr/bin/python -> python2.7*
~ ➜ sudo rm /usr/bin/python
~ ➜ sudo ln -s /usr/bin/python3 /usr/bin/python

And you’re done :slight_smile:

1 Like
16

Yes I am interested. I have some experience in deep learning.

17

I am interested to join the team

18

Any teams have been formed? I’m also interested in this =)