OpenVINO Deep Learning Workbench: A Platform for Model Optimization, Analysis and Deployment

P. 661–668.

Demidovskij A., Tugaryov A., Suvorov A., Tarkan Y., Fatekhov M., Salnikov I., Kashchikhin A., Golubenko V., Dedyukhina G., Alborova A., Palmer R., Fedorov M., Gorbachev Y.

Dramatic advances in the field of deep learning have led to superhuman results of neural models on various specialized tasks. There is an urgent need in platforms that provide performance tuning, analysis, and deployment capabilities of these models. Several pioneering platforms that have been emerging for the last five years are thoroughly analyzed and compared across numerous criteria. The OpenVINO Deep Learning Workbench (DL Workbench) is a tool designed to improve the usability and workflows for neural network performance, optimization and deployment. DL Workbench tends to play one of leading roles in terms of feature completeness across other existing platforms in the field. It provides unique capabilities of model optimization and deployment to the target hardware. By applying developer experience (DX) insights to the design of innovative deep learning solutions such as the DL Workbench, a high degree of usability can be achieved to support the complex tasks of developing highly optimized deep learning solutions for target hardware.

Keywords: artificial neural networks, deep learning, computer tools

In book

Proceedings of 2020 IEEE 32nd International Conference on Tools for Artificial Intelligence (ICTAI)

IEEE Computer Society, 2020.