Running Google MoveNet Multipose Tracking models on OpenVINO.

Overview

MoveNet Multipose Tracking on OpenVINO

Running Google MoveNet Multipose models on OpenVINO.

A convolutional neural network model that runs on RGB images and predicts human joint locations of several persons (6 max).

WIP: currently only working on CPU (not on GPU nor MYRIAD)

Demo

Full video demo here.

For MoveNet Single Pose, please visit : openvino_movenet

Install

You need OpenVINO (tested on 2021.4) and OpenCV installed on your computer and to clone/download this repository.

Run

Usage:

> python3 MovenetMPOpenvino.py -h
usage: MovenetMPOpenvino.py [-h] [-i INPUT] [--xml XML]
                            [-r {192x192,192x256,256x256,256x320,320x320,480x640,736x1280}]
                            [-t {iou,oks}] [-s SCORE_THRESHOLD] [-o OUTPUT]

optional arguments:
  -h, --help            show this help message and exit
  -i INPUT, --input INPUT
                        Path to video or image file to use as input
                        (default=0)
  --xml XML             Path to an .xml file for model
  -r {192x192,192x256,256x256,256x320,320x320,480x640,736x1280}, --res {192x192,192x256,256x256,256x320,320x320,480x640,736x1280}
  -t {iou,oks}, --tracking {iou,oks}
                        Enable tracking and specify method
  -s SCORE_THRESHOLD, --score_threshold SCORE_THRESHOLD
                        Confidence score (default=0.200000)
  -o OUTPUT, --output OUTPUT
                        Path to output video file

Examples :

  • To use default webcam camera as input :

    python3 MovenetMPOpenvino.py

  • To specify the model input resolution :

    python3 MovenetMPOpenvino.py -r 256x320

  • To enable tracking, based on Object Keypoint Similarity :

    python3 MovenetMPOpenvino.py -t keypoint

  • To use a file (video or image) as input :

    python3 MovenetMPOpenvino.py -i filename

Keypress Function
Esc Exit
space Pause
b Show/hide bounding boxes
f Show/hide FPS

Input resolution

The model input resolution (set with the '-r' or '--res' option) has an impact on the inference speed (the higher the resolution, the slower the inference) and on the size of the people that can be detected (the higher the resoltion, the smaller the size). The test below has been run on a CPU i7700k.

Resolution FPS Result
192x256 58.0 192x256
256x320 44.1 256x320
480x640 14.8 480x640
736x1280 4.5 736x1280

Tracking

The Javascript MoveNet demo code from Google proposes as an option two methods of tracking. For this repository, I have adapted this tracking code in python. You can enable the tracking with the --tracking (or -t) argument of the demo followed by iou or oks which specifies how to calculate the similarity between detections from consecutive frames :

Tracking Result
IoU Tracking IoU Tracking
OKS Tracking OKS Tracking

In the example above, we can notice several track switching in the IoU output and a track replacement (2 by 6). OKS method is doing a better job, yet it is not perfect: there is a track switching when body 3 is passing in front of body 1.

The models

The MoveNet Multipose v1 source model comes from the Tensorfow Hub: https://tfhub.dev/google/movenet/multipose/lightning/1

The model was converted by PINTO in OpenVINO IR format. Unfortunately, the OpenVINO IR MoveNet model input resolution cannot be changed dynamically, so an arbitrary list of models have been generated, each one with its dedicated input resolution. These models and others (other resolutions or precisions) are also available there: https://github.com/PINTO0309/PINTO_model_zoo/tree/main/137_MoveNet_MultiPose

Credits

You might also like...
YOLOX is a high-performance anchor-free YOLO, exceeding yolov3~v5 with ONNX, TensorRT, ncnn, and OpenVINO supported.
YOLOX is a high-performance anchor-free YOLO, exceeding yolov3~v5 with ONNX, TensorRT, ncnn, and OpenVINO supported.

Introduction YOLOX is an anchor-free version of YOLO, with a simpler design but better performance! It aims to bridge the gap between research and ind

Monocular 3D pose estimation. OpenVINO. CPU inference or iGPU (OpenCL) inference.
Monocular 3D pose estimation. OpenVINO. CPU inference or iGPU (OpenCL) inference.

human-pose-estimation-3d-python-cpp RealSenseD435 (RGB) 480x640 + CPU Corei9 45 FPS (Depth is not used) 1. Run 1-1. RealSenseD435 (RGB) 480x640 + CPU

WHENet - ONNX, OpenVINO, TFLite, TensorRT, EdgeTPU, CoreML, TFJS, YOLOv4/YOLOv4-tiny-3L
WHENet - ONNX, OpenVINO, TFLite, TensorRT, EdgeTPU, CoreML, TFJS, YOLOv4/YOLOv4-tiny-3L

HeadPoseEstimation-WHENet-yolov4-onnx-openvino ONNX, OpenVINO, TFLite, TensorRT, EdgeTPU, CoreML, TFJS, YOLOv4/YOLOv4-tiny-3L 1. Usage $ git clone htt

Python wrapper class for OpenVINO Model Server. User can submit inference request to OVMS with just a few lines of code

Python wrapper class for OpenVINO Model Server. User can submit inference request to OVMS with just a few lines of code.

Experimental Python implementation of OpenVINO Inference Engine (very slow, limited functionality). All codes are written in Python. Easy to read and modify.
Experimental Python implementation of OpenVINO Inference Engine (very slow, limited functionality). All codes are written in Python. Easy to read and modify.

PyOpenVINO - An Experimental Python Implementation of OpenVINO Inference Engine (minimum-set) Description The PyOpenVINO is a spin-off product from my

Semi-automated OpenVINO benchmark_app with variable parameters

Semi-automated OpenVINO benchmark_app with variable parameters. User can specify multiple options for any parameters in the benchmark_app and the progam runs the benchmark with all combinations of given options.

Joint detection and tracking model named DEFT, or ``Detection Embeddings for Tracking.
Joint detection and tracking model named DEFT, or ``Detection Embeddings for Tracking.

DEFT: Detection Embeddings for Tracking DEFT: Detection Embeddings for Tracking, Mohamed Chaabane, Peter Zhang, J. Ross Beveridge, Stephen O'Hara

Tracking code for the winner of track 1 in the MMP-Tracking Challenge at ICCV 2021 Workshop.

Tracking Code for the winner of track1 in MMP-Trakcing challenge This repository contains our tracking code for the Multi-camera Multiple People Track

Tracking Pipeline helps you to solve the tracking problem more easily
Tracking Pipeline helps you to solve the tracking problem more easily

Tracking_Pipeline Tracking_Pipeline helps you to solve the tracking problem more easily I integrate detection algorithms like: Yolov5, Yolov4, YoloX,

Owner
null
A Pytorch implementation of MoveNet from Google. Include training code and pre-train model.

Movenet.Pytorch Intro MoveNet is an ultra fast and accurate model that detects 17 keypoints of a body. This is A Pytorch implementation of MoveNet fro

Mr.Fire 241 Dec 26, 2022
MoveNet Single Pose on DepthAI

MoveNet Single Pose tracking on DepthAI Running Google MoveNet Single Pose models on DepthAI hardware (OAK-1, OAK-D,...). A convolutional neural netwo

null 64 Dec 29, 2022
Pose estimation with MoveNet Lightning

Pose Estimation With MoveNet Lightning MoveNet is the TensorFlow pre-trained model that identifies 17 different key points of the human body. It is th

Yash Vora 2 Jan 4, 2022
Demonstrates how to divide a DL model into multiple IR model files (division) and introduce a simplest way to implement a custom layer works with OpenVINO IR models.

Demonstration of OpenVINO techniques - Model-division and a simplest-way to support custom layers Description: Model Optimizer in Intel(r) OpenVINO(tm

Yasunori Shimura 12 Nov 9, 2022
This is a repository for a No-Code object detection inference API using the OpenVINO. It's supported on both Windows and Linux Operating systems.

OpenVINO Inference API This is a repository for an object detection inference API using the OpenVINO. It's supported on both Windows and Linux Operati

BMW TechOffice MUNICH 68 Nov 24, 2022
This is a repository for a semantic segmentation inference API using the OpenVINO toolkit

BMW-IntelOpenVINO-Segmentation-Inference-API This is a repository for a semantic segmentation inference API using the OpenVINO toolkit. It's supported

BMW TechOffice MUNICH 34 Nov 24, 2022
📚 A collection of Jupyter notebooks for learning and experimenting with OpenVINO 👓

A collection of ready-to-run Python* notebooks for learning and experimenting with OpenVINO developer tools. The notebooks are meant to provide an introduction to OpenVINO basics and teach developers how to leverage our APIs for optimized deep learning inference in their applications.

OpenVINO Toolkit 840 Jan 3, 2023
A high-performance anchor-free YOLO. Exceeding yolov3~v5 with ONNX, TensorRT, NCNN, and Openvino supported.

YOLOX is an anchor-free version of YOLO, with a simpler design but better performance! It aims to bridge the gap between research and industrial communities. For more details, please refer to our report on Arxiv.

null 7.7k Jan 6, 2023