DevMesh

Text Recognition by TestIn

Guancheng Chen

Guancheng Chen

2 1
  • 0 Collaborators

Combining AI technology to testing application in automated testing services seems to be a difficult task but has increasing needs for software providers. Thus the challenge is how to guarantee both accuracy and performance of testing workflow. We use Intel OpenVINO, oneAPI...etc. to solve this. ...learn more

Project status: Published/In Market

oneAPI, Artificial Intelligence

Intel Technologies
oneAPI, Intel vTune, OpenVINO, DevCloud

Overview / Usage

Testin is the disruptor of the traditional software testing service mode, combine AI automated real device SaaS testing. We cover 80%+ testing services of the apps provided by PRC ISVs.

We evaluated inference performance on Public Tensorflow, Intel Tensorflow (oneDNN) with custom DenseNet for OCR topology in text recognition use case.

Methodology / Approach

We used Intel® Distribution of OpenVINO™ Toolkit which supplies competitive inference speed as well as very slightly accuracy loss running on underlying Xeon® Gold processors which empower the computing power with the support of AVX-512 and oneDNN boost libraries. The Inference throughput using OpenVINO outperforms significantly 6.50x compared to Public Tensorflow on Xeon Gold 6140.

​

Technologies Used

Public Tensorflow,

Intel Tensorflow

oneDNN

VTune

OpenVINO

Comments (1)