yas-sim / openvino-auto-feature-visualization

This is a set of test program to visualize the behavior of "AUTO" plugin device switching and auto-batching feature that is newly introduced from OpenVINO 2022.1 version.

OpenVINO "AUTO" plugin / Auto-batching feature behavior visualization tool

Description:

This is a set of test programs to visualize the behavior of the "AUTO" plugin device switching and auto-batching feature that is newly introduced in the OpenVINO 2022.1 version.

• "AUTO" plugin - Web document is here.
• Auto-batching feature - Web document is here

---

How to run:

1. Install prerequisites

python3 -m pip install -r requirements.txt

2. Model Preparation (googlenet-v1-tf)

Use OMZ tools to download required DL model.

omz_downloader --list models.txt
omz_converter --list models.txt

3. Modify source code to change inference configuration.

All configuration and setting parameters are hard coded in the source code. You need to modify source code to change test configuration.

4. Run test program

• Auto-batching visualization
  This program submits asynchronous inference requests with specified interval. The program will draw the infer request submission timing on the upper row and the inference result returned timing on the lower row, connect request timing and result timing with the lines to visualize the relation of request-result timing on the frame.

python3 auto-test.py

• "AUTO" plugin device switching behavior visualization
  This program submits asynchronous infer requests with specified interval. Visualize the inference laency on a graph to visualize the inference latency fluctuation around "AUTO" plugin device switching event.

python3 auto-test-latency-graph.py

EOD