Join the public MLCommons taskforce to participate in the collaborative development of the next generation of this open-source technology!
docker:object-detection-tflite.debian-9 (v3.0.0)
Copyright: See copyright in the source repository
License: See license in the source repository
Creation date: 2019-06-01
Source: GitHub
cID: 88eef0cd8c43b68a:92486c8d2ee8ebfe

Don't hesitate to get in touch if you encounter any issues or would like to discuss this community project!
Please report if this CK component works: 1  or fails: 0 
Sign up to be notified when artifacts are shared or updated!

Description  

This CK-powered container is our attempt to provide a common API to customize, build and run AI and ML applications with different models, frameworks, libraries, datasets, compilers, formats, backends and platforms. Our on-going project is to make the onboarding process as simple as possible via this platform. Please check this CK white paper and don't hesitate to contact us if you have suggestions or feedback!

ReadMe  

MLPerf Inference - Object Detection - TFLite (Debian 9)

  1. Default image (based on Debian 9 latest)

NB: You may need to run commands below with sudo, unless you manage Docker as a non-root user.

Default image

Download

$ docker pull ctuning/object-detection-tflite.debian-9

Build

$ ck build docker:object-detection-tflite.debian-9

NB: Equivalent to:

$ cd `ck find docker:object-detection-tflite.debian-9`
$ docker build -f Dockerfile -t ctuning/object-detection-tflite.debian-9 .

Run

Object Detection (default command)

Regular NMS; 50 images
$ ck run docker:object-detection-tflite.debian-9

NB: Equivalent to:

$ docker run --rm ctuning/object-detection-tflite.debian-9 \
    "ck run program:object-detection-tflite \
        --dep_add_tags.weights=ssd-mobilenet,non-quantized --env.USE_NMS=regular \
        --dep_add_tags.dataset=coco.2017,full --env.CK_BATCH_COUNT=50 \
    "
...
Summary:
-------------------------------
All images loaded in 0.147986s
Average image load time: 0.002960s
All images detected in 5.917367s
Average detection time: 0.117556s
mAP: 0.29672520317694373
Recall: 0.3050474339529269
--------------------------------

Object Detection (custom command)

Fast NMS; 50 images
$ docker run --rm ctuning/object-detection-tflite.debian-9 \
    "ck run program:object-detection-tflite \
        --dep_add_tags.weights=ssd-mobilenet,non-quantized --env.USE_NMS=fast \
        --dep_add_tags.dataset=coco.2017,full --env.CK_BATCH_COUNT=50 \
    "
...
Summary:
-------------------------------
All images loaded in 0.146889s
Average image load time: 0.002938s
All images detected in 5.868071s
Average detection time: 0.116611s
mAP: 0.29624782705876884
Recall: 0.30501085304815917
--------------------------------
Regular NMS; 5000 images
$ docker run --rm ctuning/object-detection-tflite.debian-9 \
    "ck run program:object-detection-tflite \
        --dep_add_tags.weights=ssd-mobilenet,non-quantized --env.USE_NMS=regular \
        --dep_add_tags.dataset=coco.2017,full --env.CK_BATCH_COUNT=5000 \
    "
...
Summary:
-------------------------------
All images loaded in 14.741492s
Average image load time: 0.002948s
All images detected in 587.250183s
Average detection time: 0.117443s
mAP: 0.22349680978666922
Recall: 0.2550505369422975
--------------------------------
Fast NMS; 5000 images
$ docker run --rm ctuning/object-detection-tflite.debian-9 \
    "ck run program:object-detection-tflite \
        --dep_add_tags.weights=ssd-mobilenet,non-quantized --env.USE_NMS=fast \
        --dep_add_tags.dataset=coco.2017,full --env.CK_BATCH_COUNT=5000 \
    "
...
Summary:
-------------------------------
All images loaded in 14.953116s
Average image load time: 0.002991s
All images detected in 587.7276s
Average detection time: 0.117538s
mAP: 0.21859688835124763
Recall: 0.24801510024502602
--------------------------------

Bash

$ docker run -it --rm ctuning/object-detection-tflite.debian-9 bash

Versions  

Files  

Comments  

Please log in to add your comments!
If you notice any inapropriate content that should not be here, please report us as soon as possible and we will try to remove it within 48 hours!