Realtime prediction and voice recognition

This project will include
1. optimizing whisper’s pretrained data to Korean and Farm automation command set.
2. real time(less than 30 seconds) fine tuning of voice common(less than 5 seconds voice) under AMD Ryzen 9 7950X and GTX4090 system.
3. REST API to user’s device ( upload training data and downloading trained model)

if you don't have necessary hardware setup, you can access my on-premise server or we will set up a cloud server for this. The multi-language dataset for Korean is not provided in huggingface but you can download one from this link which need data validation;
https://aihub.or.kr/aihubdata/data/view.do?currMenu=115&topMenu=100&aihubDataSe=realm&dataSetSn=568
or download from here (tyeng0418); jupyter1.d.tyeng.com
The hugginface fine-tunning blog:
https://github.com/huggingface/blog/blob/main/fine-tune-whisper.md
The whisper blog;
https://openai.com/research/whisper
The deliverables should be available via github and implemented on the afore-mentioned server, with a documentation.

4 milestones;
1- developing the flask app
2- training data of korean whispers data and optimize it.
3- make real time prediction
4- REST API for any device