BROWSE
Detail
Conference Paper
Compatibility Technology between NNEF and ONNX using Protocol Buffer
Cited - time in 
Share 
Share
- Authors
- Jaebok Park, Kyunghee Lee, Seokjin Yoon, Seungmok Yoo, Changsik Cho
- Issue Date
- 2019-05
- Citation
- IEMEK Symposium on Embedded Technology (ISET) 2019, pp.42-44
- Publisher
- 대한임베디드공학회
- Language
- English
- Type
- Conference Paper
- Abstract
- There are various frameworks for artificial neural network deep learning development tools. However, it is implemented with individualized structuring. This requires standardization for code reuse and maintenance. Currently, neural network standardization is led by NNEF and ONNX. Previously, however, NNEF and ONNX were studying separately. Recently, efforts have been made to ensure compatibility between these two standardizations. This paper presents methods for providing compatibility between NNEF and ONNX using protocol buffer. Note that, Protocol buffers are a language-neutral, platform-neutral extensible mechanism for serializing structured data. The conversion process converts Caffe to NNEF, and then converts it back to ONNX to support compatibility.