File size: 3,172 Bytes
5aeb352 cbb82f2 312ce0c 9ebf0ef cbb82f2 e99848a d02fa41 e99848a 5aeb352 89f4096 6e42cf1 3f10107 5aeb352 9ebf0ef 5aeb352 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 |
---
license: apache-2.0
language:
- wo
metrics:
- wer
pipeline_tag: automatic-speech-recognition
model-index:
- name: wav2vec2-large-xlsr-53-wolof
results:
- task:
type: automatic-speech-recognition
name: WOLOF Speech Recognition
metrics:
- type: wer
value: 7.88
name: WER
config: load_metric("wer")
verified: true
---
## Evaluation on WOLOF Test
[](https://github.com/kingabzpro/WOLOF-ASR-Wav2Vec2)
```python
import pandas as pd
from datasets import load_dataset, load_metric,Dataset
from tqdm import tqdm
import torch
import soundfile as sf
import torchaudio
from transformers import Wav2Vec2ForCTC
from transformers import Wav2Vec2Processor
from transformers import Wav2Vec2FeatureExtractor
from transformers import Wav2Vec2CTCTokenizer
model_name = "kingabzpro/wav2vec2-large-xlsr-53-wolof"
device = "cuda"
model = Wav2Vec2ForCTC.from_pretrained(model_name).to(device)
processor = Wav2Vec2Processor.from_pretrained(model_name)
val =pd.read_csv("../input/automatic-speech-recognition-in-wolof/Test.csv")
val["path"] = "../input/automatic-speech-recognition-in-wolof/Noise Removed/tmp/WOLOF_ASR_dataset/noise_remove/"+val["ID"]+".wav"
val.rename(columns = {'transcription':'sentence'}, inplace = True)
common_voice_val = Dataset.from_pandas(val)
def speech_file_to_array_fn_test(batch):
speech_array, sampling_rate = sf.read(batch["path"])#(.wav) 16000 sample rate
batch["speech"] = speech_array
batch["sampling_rate"] = sampling_rate
return batch
def prepare_dataset_test(batch):
# check that all files have the correct sampling rate
assert (
len(set(batch["sampling_rate"])) == 1
), f"Make sure all inputs have the same sampling rate of {processor.feature_extractor.sampling_rate}."
batch["input_values"] = processor(batch["speech"], padding=True,sampling_rate=batch["sampling_rate"][0]).input_values
return batch
common_voice_val = common_voice_val.remove_columns([ "ID","age", "down_votes", "gender", "up_votes"]) # Remove columns
common_voice_val = common_voice_val.map(speech_file_to_array_fn_test, remove_columns=common_voice_val.column_names)# Applying speech_file_to_array function
common_voice_val = common_voice_val.map(prepare_dataset_test, remove_columns=common_voice_val.column_names, batch_size=8, num_proc=4, batched=True)# Applying prepare_dataset_test function
final_pred = []
for i in tqdm(range(common_voice_val.shape[0])):# Testing model on Wolof Dataset
input_dict = processor(common_voice_val[i]["input_values"], return_tensors="pt", padding=True)
logits = model(input_dict.input_values.to("cuda")).logits
pred_ids = torch.argmax(logits, dim=-1)[0]
prediction = processor.decode(pred_ids)
final_pred.append(prediction)
```
You can check my result on [Zindi](https://zindi.africa/competitions/ai4d-baamtu-datamation-automatic-speech-recognition-in-wolof/leaderboard), I got 8th rank in AI4D Baamtu Datamation - Automatic Speech Recognition in WOLOF
**Result**: 7.88 % |