Intel® Distribution of OpenVINO™ Toolkit
Community support and discussions about the Intel® Distribution of OpenVINO™ toolkit, OpenCV, and all things computer vision-related on Intel® platforms.

human pose estimation

nidhal
Beginner
274 Views

I execute the following command

python human_pose_estimation.py -m human-pose-estimation-0004.xml -i C:\Users\nidhal\Desktop\pfe_ing\ved\videoplayback.mp4 -at ae -d CPU -r 

I obtain the following result

[ INFO ] (301.91, 86.38, 0.00) (299.50, 85.58, 0.00) (321.21, 89.60, 0.01) (301.11, 85.58, 0.89) (321.21, 85.58, 0.78) (289.85, 110.49, 0.69) (330.85, 112.10, 0.68) (281.01, 142.63, 0.60) (338.89, 145.04, 0.50) (281.01, 161.12, 0.63) (338.89, 164.33, 0.62) (296.28, 169.96, 0.39) (322.01, 169.96, 0.46) (289.05, 215.76, 0.55) (331.66, 218.17, 0.57) (280.20, 260.76, 0.62) (342.91, 260.76, 0.56) | 0.50
[ INFO ] Poses:
[ INFO ] (321.21, 86.38, 0.00) (301.11, 86.38, 0.00) (321.21, 87.19, 0.01) (300.30, 85.58, 0.87) (319.60, 85.58, 0.80) (289.85, 110.49, 0.71) (330.85, 112.10, 0.68) (280.20, 141.83, 0.55) (338.89, 144.24, 0.53) (278.59, 161.92, 0.64) (340.50, 165.94, 0.59) (296.28, 169.96, 0.39) (322.01, 169.96, 0.46) (287.44, 218.97, 0.54) (331.66, 218.17, 0.56) (280.20, 260.76, 0.65) (342.91, 260.76, 0.58) | 0.50
[ INFO ] Poses:

 

you could help me to understand it

0 Kudos
1 Solution
Vladimir_Dudnik
Employee
247 Views

according to demo description, when you use `-r` command line option, demo will print to console raw detection results.

This print of raw inference results implemented by this piece of code in human_pose_estimation_demo.py file

def print_raw_results(poses, scores):
    log.info('Poses:')
    for pose, pose_score in zip(poses, scores):
        pose_str = ' '.join('({:.2f}, {:.2f}, {:.2f})'.format(p[0], p[1], p[2]) for p in pose)
        log.info('{} | {:.2f}'.format(pose_str, pose_score))

and actually, for this model, output represent 17 keypoints for each person, followed by confidence. For the details how to decode model output take a look at decoder_ae.py (for OpenVINO 2021.2 release).

Note, in OpenVINO 2021.3, which was released a few weeks ago, demo was restructured, and actual decoding was placed in model class. 

View solution in original post

2 Replies
Vladimir_Dudnik
Employee
248 Views

according to demo description, when you use `-r` command line option, demo will print to console raw detection results.

This print of raw inference results implemented by this piece of code in human_pose_estimation_demo.py file

def print_raw_results(poses, scores):
    log.info('Poses:')
    for pose, pose_score in zip(poses, scores):
        pose_str = ' '.join('({:.2f}, {:.2f}, {:.2f})'.format(p[0], p[1], p[2]) for p in pose)
        log.info('{} | {:.2f}'.format(pose_str, pose_score))

and actually, for this model, output represent 17 keypoints for each person, followed by confidence. For the details how to decode model output take a look at decoder_ae.py (for OpenVINO 2021.2 release).

Note, in OpenVINO 2021.3, which was released a few weeks ago, demo was restructured, and actual decoding was placed in model class. 

Munesh_Intel
Moderator
217 Views

Hi Nidhal,

This thread will no longer be monitored since we have provided a solution. If you need any additional information from Intel, please submit a new question.


Regards,

Munesh


Reply