From 28bb1870c82c757f5cb8feefb470855039a6f8bb Mon Sep 17 00:00:00 2001 From: Michael Freitag Date: Fri, 23 Feb 2018 11:14:25 +0100 Subject: [PATCH] Issue #3: Update documentation and sample script to account for ESC-10 having been merged with ESC-50 --- README.md | 5 ++++- samples/esc-10.sh | 3 +++ 2 files changed, 7 insertions(+), 1 deletion(-) diff --git a/README.md b/README.md index 4fc2a54..ef02bb0 100644 --- a/README.md +++ b/README.md @@ -110,8 +110,11 @@ Representation learning with auDeep is performed in several distinct stages. We use the [ESC-10 data set](https://github.com/karoldvl/ESC-10) for environmental sound classification in this guide, which contains 400 instances from 10 classes. In the command line, navigate to a directory of your choice. In the following, we will assume that any commands are executed from the directory you choose in this step. Retrieve the ESC-10 data set from Github with ``` > git clone https://github.com/karoldvl/ESC-10.git +> pushd ESC-10 +> git checkout 553c8f1743b9dba6b282e1323c3ca8fa76923448 +> popd ``` -This will store the data set in a subfolder called `ESC-10`. +This will store the data set in a subfolder called `ESC-10`. As the original ESC-10 repository has been merged with the ESC-50 repository, we have to manually checkout the correct commit. ## Extracting Spectrograms First of all, we need to extract spectrograms and some metadata from the raw audio files we downloaded during the previous step. In order to get a general overview of the audio files contained in a data set, we can use the following command. diff --git a/samples/esc-10.sh b/samples/esc-10.sh index 4ef6e18..d8162fa 100644 --- a/samples/esc-10.sh +++ b/samples/esc-10.sh @@ -57,6 +57,9 @@ fi if [ ${download_data} == "true" ]; then echo "Retrieving ESC-10 data set from https://github.com/karoldvl/ESC-10.git" git clone "https://github.com/karoldvl/ESC-10.git" ${audio_base} + pushd ${audio_base} + git checkout 553c8f1743b9dba6b282e1323c3ca8fa76923448 + popd fi # Check if auDeep has been properly installed