joel/freqsplit
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
121 Commits 1 Branch 0 Tags
cbf2b022a5de1e421128f3a36a280ad10160e461
Commit Graph

7 Commits

Author SHA1 Message Date
Joel Mathew Thomas cbf2b022a5 Replace YAMNet model for panns-inference 
YAMNet model was causing issues, as loading a pytorch framework model, when the tensorflow based YAMNet is loaded in the same environment already, caused segmentation fault
 2025-02-26 17:36:27 +05:30
Joel Mathew Thomas 3074084ac1 refactor code, use package libraries for freqsplit/preprocessing/classify 
- add additional params: sr=None and mono=False for freqsplit/input/file_reader/read_audio()
- remove loading audio using librosa in freqsplit/preprocessing/classify/classify_audio()
- add error handling, if the sr is not 16Khz
- raise RuntimeError, if YAMNet model fails
- update tests/test_preprocessing
 2025-02-25 20:13:02 +05:30
Joel Mathew Thomas 0e0e765b6e install freqsplit as a package 2025-02-25 03:10:44 +05:30
Joel Mathew Thomas 5164e2cebd rename samples/ to test_audio, and move it to tests/ 2025-01-11 03:24:05 +05:30
Joel Mathew Thomas a7176a4476 added test for resample module 2024-12-26 20:58:51 +05:30
Joel Mathew Thomas dc81f36d74 Add test case for classify_audio function 
- Implemented test case to verify audio classification with YAMNet
- Used 'cafe_crowd_talk.wav' as test file, expecting 'Speech' as the output class
 2024-12-26 01:02:10 +05:30
Joel Mathew Thomas 0ff6a12829 Implement audio normalization and trimming functions 
- Normalize audio to the range [-1, 1]
- Trim silence from audio
- Add test cases for both functions
 2024-12-24 20:59:41 +05:30