Skip to content

Latest commit

 

History

History
288 lines (240 loc) · 12.4 KB

README.md

File metadata and controls

288 lines (240 loc) · 12.4 KB

Intel(R) Deep Learning Streamer Pipeline Zoo

| Getting Started | Tasks and Pipelines | Measurement Definitions | Core Examples | Xeon Examples | Pick and Go Use Case | Advanced Examples | Pipebench Reference | Measurement Output |

The DL Streamer Pipeline Zoo is a catalog of media and media analytics pipelines optimized for Intel hardware. It includes tools for downloading pipelines and their dependencies and tools for measuring their performance.

Pipelines are organized according to the task they perform (what types of input they accept and what types of output they generate). Tasks and pipelines are defined in a platform and framework independent way to allow implementations in a variety of frameworks and for multiple platform targets.

diagram

Features Include:

Simple command line interface A single entrypoint for downloading and running media analytics pipelines along with media and model dependencies
DL Streamer Pipeline Runner Pipeline implementations and optimizations using the DL Streamer Pipeline Framework
Platform specific settings Pipeline runner settings tuned for optimal performance on different platform types (e.g. core, xeon)
Measurement Settings Settings for measuring different scenarios including single stream throughput and stream density. Settings can be customized and saved for reuse.
Containerized environment Dockerfiles, build and run scripts for launching a reproducable environment

IMPORTANT:

The DL Streamer Pipeline Zoo is provided as a set of tools for system evaluation and benchmarking and is not intended for deployment into production environments without modification.

The project is pre-production and under active development. Please expect breaking changes and use tagged versions for stable functionality.

Getting Started

Prerequisites

Docker The Pipeline Zoo requires Docker for it's build, development, and runtime environments. Please install the latest version for your platform.
bash The Pipeline Zoo build and run scripts require bash and have been tested on systems using versions greater than or equal to: GNU bash, version 4.3.48(1)-release (x86_64-pc-linux-gnu).

Installation

  1. Clone Repository
    git clone https://github.com/dlstreamer/pipeline-zoo.git pipeline-zoo
    
  2. Build Pipeline Zoo Environment
    ./pipeline-zoo/tools/docker/build.sh 
    
    Output:
    Successfully built 113352079483
    Successfully tagged media-analytics-pipeline-zoo-bench:latest
    
  3. Launch Pipeline Zoo
    ./pipeline-zoo/tools/docker/run.sh 
    

Pipline Zoo Commands

List Pipelines

Command:

pipebench list

Output:

+--------------------------------------------+-----------------------+----------------------------+------------+
| Pipeline                                   | Task                  | Models                     | Runners    |
+============================================+=======================+============================+============+
| decode-h265                                | decode-vpp            |                            | dlstreamer |
|                                            |                       |                            | mockrun    |
+--------------------------------------------+-----------------------+----------------------------+------------+
| decode-h264-bgra                           | decode-vpp            |                            | dlstreamer |
|                                            |                       |                            | mockrun    |
+--------------------------------------------+-----------------------+----------------------------+------------+
| od-h265-ssd-mobilenet-v1-coco              | object-detection      | ssd_mobilenet_v1_coco_INT8 | dlstreamer |
+--------------------------------------------+-----------------------+----------------------------+------------+
| od-h264-ssd-mobilenet-v1-coco              | object-detection      | ssd_mobilenet_v1_coco      | dlstreamer |
|                                            |                       |                            | mockrun    |
+--------------------------------------------+-----------------------+----------------------------+------------+
| oc-h265-full_frame-resnet-50-tf            | object-classification | full_frame                 | dlstreamer |
|                                            |                       | resnet-50-tf               | mockrun    |
+--------------------------------------------+-----------------------+----------------------------+------------+
| oc-h264-full_frame-resnet-50-tf            | object-classification | full_frame                 | dlstreamer |
|                                            |                       | resnet-50-tf               | mockrun    |
+--------------------------------------------+-----------------------+----------------------------+------------+
| oc-h264-ssd-mobilenet-v1-coco-resnet-50-tf | object-classification | ssd_mobilenet_v1_coco      | dlstreamer |
|                                            |                       | resnet-50-tf               | mockrun    |
+--------------------------------------------+-----------------------+----------------------------+------------+
| oc-h265-ssd-mobilenet-v1-coco-resnet-50-tf | object-classification | ssd_mobilenet_v1_coco      | dlstreamer |
|                                            |                       | resnet-50-tf               | mockrun    |
+--------------------------------------------+-----------------------+----------------------------+------------+

Download Pipeline

Command:

pipebench download od-h264-ssd-mobilenet-v1-coco

Example Output Tree:

- pipeline-zoo/
+ doc/
+ media/
+ models/
+ pipelines/
+ runners/
+ tools/
- workspace/
 - od-h264-ssd-mobilenet-v1-coco/
   - media/
     - video/
       + Pexels-Videos-1388365/
       + person-bicycle-car-detection/
   - models/
     - ssd_mobilenet_v1_coco/
       + FP16/
       + FP32/
       + ssd_mobilenet_v1_coco_2018_01_28/
     - ssd_mobilenet_v1_coco_INT8/
       + FP16-INT8/
   - runners/
     + dlstreamer/
     + mockrun/
   README.md
   dlstreamer.core.runner-settings.yml
   dlstreamer.density.core.runner-settings.yml
   dlstreamer.density.dgpu.runner-settings.yml
   dlstreamer.density.runner-settings.yml
   dlstreamer.density.xeon.runner-settings.yml
   dlstreamer.dgpu.runner-settings.yml
   dlstreamer.runner-settings.yml
   dlstreamer.xeon.runner-settings.yml
   media.list.yml
   mockrun.runner-settings.yml
   models.list.yml
   od-h264-ssd-mobilenet-v1-coco.pipeline.yml

Measure Single Stream Throughput

Command:

pipebench run od-h264-ssd-mobilenet-v1-coco

Example Output:

 Pipeline:
	od-h264-ssd-mobilenet-v1-coco

 Runner:
	dlstreamer
 	dlstreamer.runner-settings.yml

 Media:
	video/person-bicycle-car-detection

 Measurement:
	throughput
 	throughput.measurement-settings.yml

 Output Directory:
	/home/pipeline-zoo/workspace/od-h264-ssd-mobilenet-v1-coco/measurements/throughput/dlstreamer/run-0000

========================================================================
Iteration   Streams  Processes    Minimum   Average   Maximum      Total
========================================================================
     0000      0000       0001     0.0000    0.0000    0.0000     0.0000
======================================================================== 

========================================================================
Iteration   Streams  Processes    Minimum   Average   Maximum      Total
========================================================================
     0000      0001       0001   130.3469  130.3469  130.3469   130.3469
======================================================================== 

========================================================================
Iteration   Streams  Processes    Minimum   Average   Maximum      Total
========================================================================
     0000      0001       0001   128.9403  128.9403  128.9403   128.9403
======================================================================== 

========================================================================
Iteration   Streams  Processes    Minimum   Average   Maximum      Total
========================================================================
     0000      0001       0001   129.5578  129.5578  129.5578   129.5578
======================================================================== 

...
<SNIP>
...

========================================================================
Iteration   Streams  Processes    Minimum   Average   Maximum      Total
========================================================================
     0000      0001       0001   126.2640  126.2640  126.2640   126.2640
======================================================================== 

========================================================================
Iteration   Streams  Processes    Minimum   Average   Maximum      Total
========================================================================
     0000      0001       0001   125.8236  125.8236  125.8236   125.8236
======================================================================== 

Pipeline                       Runner      Streams: 1
-----------------------------  ----------  ---------------------------------------------------------
od-h264-ssd-mobilenet-v1-coco  dlstreamer  Min: 125.8236 Max: 125.8236 Avg: 125.8236 Total: 125.8236

Measure Stream Density

Command:

 pipebench run --measure density od-h264-ssd-mobilenet-v1-coco

Example Output:

 Pipeline:
	od-h264-ssd-mobilenet-v1-coco

 Runner:
	dlstreamer
 	dlstreamer.density.runner-settings.yml

 Media:
	video/person-bicycle-car-detection

 Measurement:
	density
 	density.measurement-settings.yml

 Output Directory:
	/home/pipeline-zoo/workspace/od-h264-ssd-mobilenet-v1-coco/measurements/density/dlstreamer/run-0000

========================================================================
Iteration   Streams  Processes    Minimum   Average   Maximum      Total
========================================================================
      PRE      0001       0001   121.7170  121.7170  121.7170   121.7170
======================================================================== 

========================================================================
Iteration   Streams  Processes    Minimum   Average   Maximum      Total
========================================================================
      PRE      0001       0001   128.3342  128.3342  128.3342   128.3342
======================================================================== 

...
<SNIP>
...

========================================================================
Iteration   Streams  Processes    Minimum   Average   Maximum      Total
========================================================================
     0001      0003       0003    30.0000   30.0038   30.0110    90.0115
======================================================================== 

========================================================================
Iteration   Streams  Processes    Minimum   Average   Maximum      Total
========================================================================
     0001      0003       0003    29.9868   29.9959   30.0115    89.9878
======================================================================== 

Pipeline                       Runner      Streams: 4                                              Streams: 3
-----------------------------  ----------  ------------------------------------------------------  -----------------------------------------------------
od-h264-ssd-mobilenet-v1-coco  dlstreamer  Min: 28.4167 Max: 28.5507 Avg: 28.4844 Total: 113.9374  Min: 29.9868 Max: 30.0115 Avg: 29.9959 Total: 89.9878