Installing required software

Make sure you are connected to the Internet.

Updating macOS

  1. Ensure you are running the latest version of macOS. You can check the latest versions of macOS on the Apple website. 

Installing the Procyon AI Computer Vision Benchmark application

  1. Install Procyon:
    Procyon-<version>-ai-computer-vision-benchmark-macos.pkg
  2. Follow the prompts to complete the installation. 

Note:  
the UL Procyon AI Computer Vision Benchmark is installed to the ‘/Library/UL/Procyon/AIComputerVision’ folder. 

Registering your Procyon license 

  1. Open Terminal. 
  2. Go to installation folder:  
    cd /Library/UL/Procyon/AIComputerVision
  3. Register your license:
    ./UL_Procyon --register <license key>

Running the AI Computer Vision Benchmark 

Once the benchmark is installed and permissions configured properly, it can be run as follows: 

  1. Open Terminal. 
  2. In Terminal, navigate to the installation folder with:
    cd /Library/UL/Procyon/AIComputerVision
  3. Run the Computer Vision benchmark:
    ./UL_Procyon -d ai_computer_vision_coreml_<device type>_<precision>.def -o “<output path>/<output file>.procyon-result” 
    Example:
    (Please note: "ProcyonTestUser/Procyonresults" is an example folder. attempting to export results to a folder that does not exist will mean no results file is created.)
    ./UL_Procyon -d ai_computer_vision_coreml_all_fp16.def 
    -o “/Users/ProcyonTestUser/Procyonresults/AIComputerVisionResult.procyon-result” 

Notes:

  1. Unlike the AI Computer Vision Benchmark on Windows, we have already created individual definition files for each configuration of device type and precision. The complete list can be found below
    ai_computer_vision_coreml_all_fp16.def
ai_computer_vision_coreml_all_fp32.def
ai_computer_vision_coreml_all_integer.def
ai_computer_vision_coreml_cpu_and_gpu_fp16.def
ai_computer_vision_coreml_cpu_and_gpu_fp32.def
ai_computer_vision_coreml_cpu_and_gpu_integer.def
ai_computer_vision_coreml_cpu_and_ne_fp16.def
ai_computer_vision_coreml_cpu_and_ne_fp32.def
ai_computer_vision_coreml_cpu_and_ne_integer.def
ai_computer_vision_coreml_cpu_fp16.def
ai_computer_vision_coreml_cpu_fp32.def
ai_computer_vision_coreml_cpu_integer.def
  2. If the output path has whitespace, it must be escaped (note quotation marks and backslash).  
    Example:
    for output folder “/Users/Steve/result files/“ :   
    -o “/Users/<username>/result\ files/AIComputerVisionResult.procyon-result” 

Viewing your results

Exporting with command line switch

Option 1:

  1. Use -export-csv or -export-xml command line switch to export results as an XML or CSV file:  
    ./UL_Procyon -d ai_computer_vision_coreml_<device type>_<precision>.def -o <Export File Name>.zip --export-csv <CSV File Name>.csv
    Example:
    ./UL_Procyon -d ai_computer_vision_coreml_all_fp16.def -o ZipAIComputerVisionResult.zip --export-csv AIComputerVisionResult.csv

Option 2:

  1. Use the -export-simple-CSV command line switch to export results as a single CSV file:  
    ./UL_Procyon -d ai_computer_vision_coreml_<device type>_<precision>.def -o <Export File Name>.zip -l 2 --export-simple-csv <CSV File Name>.csv
    Example:
    ./UL_Procyon -d ai_computer_vision_coreml_all_fp16.def -o ZipAIComputerVisionResult.zip -l 2 --export-simple-csv AIComputerVisionResult.csv

Extracting Result.xml from the .procyon-result file.

(This is essentially the same as using –export-xml command) 

  1. Unzip the .procyon-result file to extract the contents. Your benchmark results are available in the ‘Result.xml’ file.  

Once you have the CSV or XML file, you can process the data with your preferred tools.


Uninstalling UL Procyon AI Computer Vision Benchmark 

The benchmark can be uninstalled as follows: 

  1. Open Terminal. 
  2. In Terminal, navigate to the installation folder with:
    cd /Library/UL/Procyon/AIComputerVision
  3. Run the uninstallation script:
    sudo ./uninstall.sh

Note:   

It is not recommended to uninstall just by removing the installation folder, as it will leave the Procyon app installed in the operating system registers. This can cause issues when installing Procyon again later.