search

MrYANG23 / Audio2BS-AIWIN2022-16th

2 stars 0 forks source link

about dataset #1

Open pipijiev12 opened 10 months ago

pipijiev12 commented 10 months ago

In addition to the aiwin2022 data set, are there other data sets that can be trained? ?

MrYANG23 commented 10 months ago

Yes, this code can also train on other datasets and achieved a top 7 result in this competition (https://aicontest.msxf.com/vision)

pipijiev12 commented 10 months ago

Yes, this code can also train on other datasets and achieved a top 7 result in this competition (https://aicontest.msxf.com/vision) It's a pity that the competition has ended. I can't apply to obtain the data. Or can you provide relevant software for me to produce the data?

MrYANG23 commented 10 months ago

I can't share this dataset, but any dataset in the form of speech-to-blendshape coefficient data can be used to train this model

pipijiev12 commented 10 months ago

I can't share this dataset, but any dataset in the form of speech-to-blendshape coefficient data can be used to train this model Do I only need the following two types of files: data preprocessing and pretraining model? One is a wav file and the other is a json file containing bone parameters.

a2f_export_bsweight_chinese_voice_female_p01_neutral.json

yonzhao commented 3 months ago

Hi MrYang

I 'am so sorry to bother you. I have three question to ask you.

  1. Could you describe the contest data, one person voice for 30 mins?
  2. I trained my model, it's bad for the close sound such as 'b p m f ', do you meet the issue, or could you give me some solution?
  3. For the different voice, your model's generalization is good ? I'am looking forward to your reply.