human-pose-estimation-3d-0001

Use Case and High-Level Description

Multi-person 3D human pose estimation model based on the Lightweight OpenPose and Single-Shot Multi-Person 3D Pose Estimation From Monocular RGB papers.

Example

Specification

Metric

Value

MPJPE (mm)

100.45

GFlops

18.998

MParams

5.074

Source framework

PyTorch*

Inputs

Image, name: data, shape: 1, 3, 256, 448 in the B, C, H, W format, where:

  • B - batch size

  • C - number of channels

  • H - image height

  • W - image width

Expected color order is BGR.

Outputs

The net outputs are three blobs:

  1. Name: features, shape: 1, 57, 32, 56 - coordinates in 3D space.

  2. Name: heatmaps, shape: 1, 19, 32, 56 - keypoint heatmaps.

  3. Name: pafs, shape: 1, 38, 32, 56 - keypoint pairwise relations (part affinity fields).

Download a Model and Convert it into OpenVINO™ IR Format

You can download models and if necessary convert them into OpenVINO™ IR format using the Model Downloader and other automation tools as shown in the examples below.

An example of using the Model Downloader:

omz_downloader --name <model_name>

An example of using the Model Converter:

omz_converter --name <model_name>

Demo usage

The model can be used in the following demos provided by the Open Model Zoo to show its capabilities: