search

Daniil-Osokin / lightweight-human-pose-estimation-3d-demo.pytorch

Real-time 3D multi-person pose estimation demo in PyTorch. OpenVINO backend can be used for fast inference on CPU.
Apache License 2.0
653 stars 137 forks source link

poses_3d description #70

Closed AndreyGalitsin closed 3 years ago

AndreyGalitsin commented 3 years ago

Hi! I have run a demo file and I have a NN's output with the following shape (1,19, 3). Can you explain please, what does each of 19 points mean (where are eyes, nose, arms and so on)?

Daniil-Osokin commented 3 years ago

Hi! Will #64, #28 help?

AndreyGalitsin commented 3 years ago

Yes, thank you. I was looking for this line!

AndreyGalitsin commented 3 years ago

Sorry, another question. As I understand, you don't use the 2nd point in a 3d Model. It will be added in the future. Am I right? And also, I ran this detector and got the following results: 0 point: [-68.575226 108.350235 97.04121 ] 1 point: [-51.340336 113.00133 95.14712 ]

2 point: [-95.47389 100.40167 61.071163]

3 point: [-72.04748 120.715416 98.79489 ] 4 point: [-76.18896 123.19751 77.983894] 5 point: [-59.074917 125.94555 70.73402 ] 6 point: [-90.563354 109.99237 60.912197] 7 point: [-79.10259 123.149666 37.453236] 8 point: [-84.571205 126.454956 6.46569 ] 9 point: [-67.154366 94.222206 96.09957 ] 10 point: [-67.52125 87.26843 75.11487] 11 point: [-54.099243 96.905594 72.4959 ] 12 point: [-74.22948 95.94762 62.804024] 13 point: [-46.138985 103.71267 46.099907] 14 point: [-47.302933 108.139404 12.033979] 15 point: [-47.294724 114.92059 91.96468 ] 16 point: [-48.200966 112.0552 81.89803 ] 17 point: [-49.961666 113.7537 96.478874] 18 point: [-54.311226 109.55264 100.6151 ]

X coordinate is negative. Is it normal? And tell me please, where is the origin of coordinates?

Daniil-Osokin commented 3 years ago

Hey, Andrey! Have you seen #57, will it help?

Daniil-Osokin commented 3 years ago

Hope, it helped.