killua
/
TakwayDisplayPlatform
1
0
Fork 1
Code Issues Pull Requests 1 Packages Projects Releases 1 Wiki Activity

WIP: bing-patch-1 #2

Draft
bing wants to merge 3 commits from bing/TakwayDisplayPlatform:bing-patch-1 into master
pull from: bing/TakwayDisplayPlatform:bing-patch-1
merge into: killua:master
Conversation 0 Commits 3 Files Changed 1 +53 -36
1 changed files with 53 additions and 36 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files
Showing only changes of commit 25b99502fd - Show all commits

89
utils/emotion2vec_utils.py
View File

@ -1,37 +1,54 @@
from modelscope.pipelines import pipeline from modelscope.pipelines import pipeline
from modelscope.utils.constant import Tasks from modelscope.utils.constant import Tasks
import json import json
model_args = { model_args = {
"task": Tasks.emotion_recognition, "task": Tasks.emotion_recognition,
"model": "iic/emotion2vec_base_finetuned" # Alternative: iic/emotion2vec_plus_seed, iic/emotion2vec_plus_base, iic/emotion2vec_plus_large and iic/emotion2vec_base_finetuned "model": "iic/emotion2vec_base_finetuned" # Alternative: iic/emotion2vec_plus_seed, iic/emotion2vec_plus_base, iic/emotion2vec_plus_large and iic/emotion2vec_base_finetuned
# "device": 不用指定，"device" 默认为gpu # "device": 不用指定，"device" 默认为gpu
} }
class EmotionRecognition: class EmotionRecognition:
def __init__(self) -> None: """
self.initialize(model_args=model_args) 9-class emotion
0: angry
# 初始化模型 1: disgusted
def initialize(self, model_args=model_args): 2: fearful
self.inference_pipeline = pipeline(**model_args) 3: happy
4: neutral
def emotion_recognition(self, 5: other
audio:bytes, 6: sad
granularity="utterance", # 中间特征的维度，"utterance": [*768], "frame": [T*768] 7: surprised
extract_embedding=False, # 是否保留提取到的特征，False表示不保留中间特征，只保留最终结果 8: unknown
output_dir="./outputs" # 中间特征的保存位置(extract_embedding为true时有效) return :
): [{"emotion": "angry", "weight": },
rec_result = self.inference_pipeline(audio, granularity=granularity, extract_embedding=extract_embedding, output_dir=output_dir) {"emotion": "disgusted", "weight": },
...
# 保存结果 ]
json_list = [] """
for emotion, score in zip(rec_result[0]["labels"], rec_result[0]["scores"]): def __init__(self) -> None:
json_list.append({"emotion": emotion.split("/")[-1], "weight": round(score,4)}) self.initialize(model_args=model_args)
recognize_result = json.dumps(json_list)
return recognize_result # 初始化模型
def initialize(self, model_args=model_args):
if __name__ == "__main__": self.inference_pipeline = pipeline(**model_args)
model = EmotionRecognition()
recognize_result = model.emotion_recognition("https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/asr_example_zh.wav") def emotion_recognition(self,
audio:bytes,
granularity="utterance", # 中间特征的维度，"utterance": [*768], "frame": [T*768]
extract_embedding=False, # 是否保留提取到的特征，False表示不保留中间特征，只保留最终结果
output_dir="./outputs" # 中间特征的保存位置(extract_embedding为true时有效)
):
rec_result = self.inference_pipeline(audio, granularity=granularity, extract_embedding=extract_embedding, output_dir=output_dir)
# 保存结果
json_list = []
for emotion, score in zip(rec_result[0]["labels"], rec_result[0]["scores"]):
json_list.append({"emotion": emotion.split("/")[-1], "weight": round(score,4)})
recognize_result = json.dumps(json_list)
return recognize_result
if __name__ == "__main__":
model = EmotionRecognition()
recognize_result = model.emotion_recognition("https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/asr_example_zh.wav")
print(recognize_result) print(recognize_result)