experiments
/
STT
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Also remove samples with noise 

X-DeepSpeech: NOBUILD
This commit is contained in:
Reuben Morais 2019-09-11 09:02:21 +00:00 committed by GitHub
parent 90c2acd810
commit fcb9bf6d9f
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 4 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
bin/import_magicdata.py
View File

@ -66,6 +66,10 @@ def preprocess_data(folder_with_archives, target_dir):
df = df[durations <= 10.0] df = df[durations <= 10.0]
print('Trimming {} samples > 10 seconds'.format((durations > 10.0).sum())) print('Trimming {} samples > 10 seconds'.format((durations > 10.0).sum()))
with_noise = df['transcript'].str.contains(r'\[(FIL|SPK)\]')
df = df[~with_noise]
print('Trimming {} samples with noise ([FIL] or [SPK])'.format(sum(with_noise)))
dest_csv = os.path.join(target_dir, 'magicdata_{}.csv'.format(subset)) dest_csv = os.path.join(target_dir, 'magicdata_{}.csv'.format(subset))
print('Saving {} set into {}...'.format(subset, dest_csv)) print('Saving {} set into {}...'.format(subset, dest_csv))
df.to_csv(dest_csv, index=False) df.to_csv(dest_csv, index=False)