chrisjay commited on
Commit
cd48f69
1 Parent(s): 5f193ba

added updates

Browse files
README.md ADDED
@@ -0,0 +1,75 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ tags:
4
+ - afro-digits-speech
5
+ datasets:
6
+ - crowd-speech-africa
7
+ metrics:
8
+ - accuracy
9
+ model-index:
10
+ - name: afrospeech-wav2vec-ibo
11
+ results:
12
+ - task:
13
+ name: Audio Classification
14
+ type: audio-classification
15
+ dataset:
16
+ name: Afro Speech
17
+ type: chrisjay/crowd-speech-africa
18
+ args: no
19
+ metrics:
20
+ - name: Validation Accuracy
21
+ type: accuracy
22
+ value: 1.0
23
+ ---
24
+
25
+
26
+ # afrospeech-wav2vec-ibo
27
+
28
+ This model is a fine-tuned version of [facebook/wav2vec2-base](https://huggingface.co/facebook/wav2vec2-base) on the [crowd-speech-africa](https://huggingface.co/datasets/chrisjay/crowd-speech-africa), which was a crowd-sourced dataset collected using the [afro-speech Space](https://huggingface.co/spaces/chrisjay/afro-speech). It achieves the following results on the [validation set](VALID_igbo_ibo_audio_data.csv):
29
+
30
+ - F1: 1.0
31
+ - Accuracy: 1.0
32
+
33
+ The confusion matrix below helps to give a better look at the model's performance across the digits. Through it, we can see the precision and recall of the model as well as other important insights.
34
+
35
+ ![confusion matrix](afrospeech-wav2vec-ibo_confusion_matrix_VALID.png)
36
+
37
+
38
+ ## Training and evaluation data
39
+
40
+ The model was trained on a mixed audio data from Igbo (`ibo`).
41
+
42
+ - Size of training set: 109
43
+ - Size of validation set: 28
44
+
45
+ Below is a distribution of the dataset (training and valdation)
46
+
47
+ ![digits-bar-plot-for-afrospeech](digits-bar-plot-for-afrospeech-wav2vec-ibo.png)
48
+
49
+
50
+ ### Training hyperparameters
51
+
52
+ The following hyperparameters were used during training:
53
+ - learning_rate: 3e-05
54
+ - train_batch_size: 64
55
+ - eval_batch_size: 64
56
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
57
+ - num_epochs: 150
58
+
59
+ ### Training results
60
+
61
+ | Training Loss | Epoch | Validation Accuracy |
62
+ |:-------------:|:-----:|:--------:|
63
+ | 0.1415 | 1 | 1.0 |
64
+ | 0.0241 | 50 | 1.0 |
65
+ | 0.0019 | 100 | 0.929 |
66
+ | 0.0012 | 150 | 0.892 |
67
+
68
+
69
+
70
+ ### Framework versions
71
+
72
+ - Transformers 4.21.3
73
+ - Pytorch 1.12.0
74
+ - Datasets 1.14.0
75
+ - Tokenizers 0.12.1
VALID_igbo_ibo_audio_data.csv ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ audio_path,transcript,lang,lang_code,gender,age,country,accent
2
+ AUDIO_HOMEPATH/data/5g8jjWfDIyVQS1yTa3xqGALSDz4vq2dl/audio.wav,0,igbo,ibo,,,,
3
+ AUDIO_HOMEPATH/data/Nqfv2geltUbENlNHJFDdvzTJNPFHuT5z/audio.wav,4,igbo,ibo,Male,,,
4
+ AUDIO_HOMEPATH/data/rJqbQez84emINUAWMOD45hmiYlc0GSlP/audio.wav,5,igbo,ibo,Female,24.0,Nigeria,
5
+ AUDIO_HOMEPATH/data/prIkLOlvFzPpHl81wrxHcEUmt2ZHhSJO/audio.wav,0,igbo,ibo,,,,
6
+ AUDIO_HOMEPATH/data/cN25SF7c3nWsvYNTfAyyBspa6nuiLsyT/audio.wav,8,igbo,ibo,Female,24.0,Nigeria,
7
+ AUDIO_HOMEPATH/data/gyzTJbN8MiDZQMg7s6UypvCTY6P2x61D/audio.wav,5,igbo,ibo,Male,25.0,Nigeria,Igbo
8
+ AUDIO_HOMEPATH/data/cmuacrRnKyffuq9Qh1ibzYbdXh2wvoQg/audio.wav,6,igbo,ibo,,,,
9
+ AUDIO_HOMEPATH/data/AKdDTYyPqgBpQiF3neHd68zVFEAJr8uo/audio.wav,0,igbo,ibo,,,,
10
+ AUDIO_HOMEPATH/data/VkfjxzC8TpDMzSy1lzKW8KIR3cBCTgX9/audio.wav,6,igbo,ibo,Male,28.0,Nigeria,
11
+ AUDIO_HOMEPATH/data/nNIo0e5ibf2zQIFBPxs2TfC5qSOhiHGJ/audio.wav,1,igbo,ibo,Male,25.0,Nigeria,Igbo
12
+ AUDIO_HOMEPATH/data/XLQjgmS4JxvuIO5mkSTZrFB4dVnXwQaM/audio.wav,1,igbo,ibo,Male,28.0,Nigeria,
13
+ AUDIO_HOMEPATH/data/V9lvNLXsgp8aY0xeMD5xxLh6KHWiLPCe/audio.wav,4,igbo,ibo,Female,25.0,Nigeria,
14
+ AUDIO_HOMEPATH/data/06M3yvaKEmEFkcuDwVy3aBrKo9TB5IVY/audio.wav,0,igbo,ibo,Female,25.0,Nigeria,
15
+ AUDIO_HOMEPATH/data/k5YpfgDC0dNOAOgAWMMK7g32oXKwCLkr/audio.wav,8,igbo,ibo,Male,25.0,Nigeria,Igbo
16
+ AUDIO_HOMEPATH/data/PMAqmmjgmKVpoI5fCW10111DGD8Dew0H/audio.wav,2,igbo,ibo,Female,28.0,Nigeria,Igbo
17
+ AUDIO_HOMEPATH/data/EKXa4vADiG44Q4gKZLVMgCw2n2y2Odbt/audio.wav,8,igbo,ibo,Female,25.0,Nigeria,
18
+ AUDIO_HOMEPATH/data/0902UDIFwFBgM40EM2Nmuy0KvGIGRQ9p/audio.wav,5,igbo,ibo,Male,26.0,Nigeria,
19
+ AUDIO_HOMEPATH/data/UEN6f6kTqDO3AA04ZWVHr22u3nalfOS9/audio.wav,0,igbo,ibo,,,,
20
+ AUDIO_HOMEPATH/data/F2ANjx0yN60UYCKuLNzg4P9tdZFt6PnW/audio.wav,3,igbo,ibo,Female,25.0,Nigeria,
21
+ AUDIO_HOMEPATH/data/eFJPnuqHhn6RnlRs0jWvR6cT78Kg5YTT/audio.wav,3,igbo,ibo,,29.0,Nigeria,Afikpo
22
+ AUDIO_HOMEPATH/data/sAymElmCNJhWOKUoovIhgRnfZCRSwA22/audio.wav,1,igbo,ibo,Female,25.0,Nigeria,
23
+ AUDIO_HOMEPATH/data/9Es6Xy9ao4nThotFbxzP2d79DMlNZu65/audio.wav,7,igbo,ibo,Male,26.0,Nigeria,
24
+ AUDIO_HOMEPATH/data/bqy4vH5S1W7YowNwTBQkAXe0IOBjxc9K/audio.wav,5,igbo,ibo,Female,28.0,Nigeria,Igbo
25
+ AUDIO_HOMEPATH/data/0xDhsm0fDV30QC5g3oVPOsK3pd4quzcq/audio.wav,6,igbo,ibo,Female,28.0,Nigeria,Igbo
26
+ AUDIO_HOMEPATH/data/33EdKsLcgmmYxQh9nePHb8eqm93jU3GV/audio.wav,4,igbo,ibo,Male,29.0,Nigeria,Igbo
27
+ AUDIO_HOMEPATH/data/oXo6SpSEJBcIV2shN9th6w0jIRzEcBZm/audio.wav,0,igbo,ibo,,,,
28
+ AUDIO_HOMEPATH/data/H30cvWGgwJNKnZ9ALTksPquyeeCt74ev/audio.wav,6,igbo,ibo,,,,
29
+ AUDIO_HOMEPATH/data/Czp0oniOFjAcn4owZtu98Hk8ZE5fid9C/audio.wav,3,igbo,ibo,Female,24.0,Nigeria,
afrospeech-wav2vec-ibo_METRICS_VALID.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"acc": 1.0, "f1": 1.0}
afrospeech-wav2vec-ibo_confusion_matrix_VALID.png ADDED
digits-bar-plot-for-afrospeech-wav2vec-ibo.png ADDED