Direct Multi-view Multi-person 3D Human Pose Estimation

Related tags

Miscellaneousmvp
Overview

Implementation of NeurIPS-2021 paper: Direct Multi-view Multi-person 3D Human Pose Estimation

[paper] [video-YouTube, video-Bilibili] [slides]

This is the official implementation of our NeurIPS-2021 work: Multi-view Pose Transformer (MvP). MvP is a simple algorithm that directly regresses multi-person 3D human pose from multi-view images.

Framework

mvp_framework

Example Result

mvp_framework

Reference

@article{wang2021mvp,
  title={Direct Multi-view Multi-person 3D Human Pose Estimation},
  author={Tao Wang and Jianfeng Zhang and Yujun Cai and Shuicheng Yan and Jiashi Feng},
  journal={Advances in Neural Information Processing Systems},
  year={2021}
}

1. Installation

  1. Set the project root directory as ${POSE_ROOT}.
  2. Install all the required python packages (with requirements.txt).
  3. compile deformable operation for projective attention.
cd ./models/ops
sh ./make.sh

2. Data and Pre-trained Model Preparation

2.1 CMU Panoptic

Please follow VoxelPose to download the CMU Panoptic Dataset and PoseResNet-50 pre-trained model.

The directory tree should look like this:

${POSE_ROOT}
|-- models
|   |-- pose_resnet50_panoptic.pth.tar
|-- data
|   |-- panoptic
|   |   |-- 16060224_haggling1
|   |   |   |-- hdImgs
|   |   |   |-- hdvideos
|   |   |   |-- hdPose3d_stage1_coco19
|   |   |   |-- calibration_160224_haggling1.json
|   |   |-- 160226_haggling1
|   |   |-- ...

2.2 Shelf/Campus

Please follow VoxelPose to download the Shelf/Campus Dataset.

Due to the limited and incomplete annotations of the two datasets, we use psudo ground truth 3D pose generated from VoxelPose to train the model, we expect mvp would perform much better with absolute ground truth pose data.

Please use voxelpose or other methods to generate psudo ground truth for the training set, you can also use our generated psudo GT: psudo_gt_shelf. psudo_gt_campus. psudo_gt_campus_fix_gtmorethanpred.

Due to the small dataset size, we fine-tune Panoptic pre-trained model to Shelf and Campus. Download the pretrained MvP on Panoptic from model_best_5view and model_best_3view_horizontal_view or model_best_3view_2horizon_1lookdown

The directory tree should look like this:

${POSE_ROOT}
|-- models
|   |-- model_best_5view.pth.tar
|   |-- model_best_3view_horizontal_view.pth.tar
|   |-- model_best_3view_2horizon_1lookdown.pth.tar
|-- data
|   |-- Shelf
|   |   |-- Camera0
|   |   |-- ...
|   |   |-- Camera4
|   |   |-- actorsGT.mat
|   |   |-- calibration_shelf.json
|   |   |-- pesudo_gt
|   |   |   |-- voxelpose_pesudo_gt_shelf.pickle
|   |-- CampusSeq1
|   |   |-- Camera0
|   |   |-- Camera1
|   |   |-- Camera2
|   |   |-- actorsGT.mat
|   |   |-- calibration_campus.json
|   |   |-- pesudo_gt
|   |   |   |-- voxelpose_pesudo_gt_campus.pickle
|   |   |   |-- voxelpose_pesudo_gt_campus_fix_gtmorethanpred_case.pickle

2.3 Human3.6M dataset

Please follow CHUNYUWANG/H36M-Toolbox to prepare the data.

2.4 Full Directory Tree

The data and pre-trained model directory tree should look like this, you can only download the Panoptic dataset and PoseResNet-50 for reproducing the main MvP result and ablation studies:

${POSE_ROOT}
|-- models
|   |-- pose_resnet50_panoptic.pth.tar
|   |-- model_best_5view.pth.tar
|   |-- model_best_3view_horizontal_view.pth.tar
|   |-- model_best_3view_2horizon_1lookdown.pth.tar
|-- data
|   |-- pesudo_gt
|   |   |-- voxelpose_pesudo_gt_shelf.pickle
|   |   |-- voxelpose_pesudo_gt_campus.pickle
|   |   |-- voxelpose_pesudo_gt_campus_fix_gtmorethanpred_case.pickle
|   |-- panoptic
|   |   |-- 16060224_haggling1
|   |   |   |-- hdImgs
|   |   |   |-- hdvideos
|   |   |   |-- hdPose3d_stage1_coco19
|   |   |   |-- calibration_160224_haggling1.json
|   |   |-- 160226_haggling1
|   |   |-- ...
|   |-- Shelf
|   |   |-- Camera0
|   |   |-- ...
|   |   |-- Camera4
|   |   |-- actorsGT.mat
|   |   |-- calibration_shelf.json
|   |   |-- pesudo_gt
|   |   |   |-- voxelpose_pesudo_gt_shelf.pickle
|   |-- CampusSeq1
|   |   |-- Camera0
|   |   |-- Camera1
|   |   |-- Camera2
|   |   |-- actorsGT.mat
|   |   |-- calibration_campus.json
|   |   |-- pesudo_gt
|   |   |   |-- voxelpose_pesudo_gt_campus.pickle
|   |   |   |-- voxelpose_pesudo_gt_campus_fix_gtmorethanpred_case.pickle
|   |-- HM36

3. Training and Evaluation

The evaluation result will be printed after every epoch, the best result can be found in the log.

3.1 CMU Panoptic dataset

We train and validate on the five selected camera views. We trained our models on 8 GPUs and batch_size=1 for each GPU, note the total iteration per epoch should be 3205, if not, please check your data.

python -m torch.distributed.launch --nproc_per_node=8 --use_env run/train_3d.py --cfg configs/panoptic/best_model_config.yaml

Pre-trained models

Datasets AP25 AP25 AP25 AP25 MPJPE pth
Panoptic 92.3 96.6 97.5 97.7 15.8 here

3.1.1 Ablation Experiments

You can find several ablation experiment configs under ./configs/panoptic/, for example, removing RayConv:

python -m torch.distributed.launch --nproc_per_node=8 --use_env run/train_3d.py --cfg configs/panoptic/ablation_remove_rayconv.yaml

3.2 Shelf/Campus datasets

As shelf/campus are very small dataset with incomplete annotation, we finetune pretrained MvP with pseudo ground truth 3D pose extracted with VoxelPose, we expect more accurate GT would help MvP achieve much higher performance.

python -m torch.distributed.launch --nproc_per_node=8 --use_env run/train_3d.py --cfg configs/shelf/mvp_shelf.yaml

Pre-trained models

Datasets Actor 1 Actor 2 Actor 2 Average pth
Shelf 99.3 95.1 97.8 97.4 here
Campus 98.2 94.1 97.4 96.6 here

3.3 Human3.6M dataset

MvP also applies to the naive single-person setting, with dataset like Human3.6, to come

python -m torch.distributed.launch --nproc_per_node=8 --use_env run/train_3d.py --cfg configs/h36m/mvp_h36m.yaml

4. Evaluation Only

To evaluate a trained model, pass the config and model pth:

python -m torch.distributed.launch --nproc_per_node=8 --use_env run/validate_3d.py --cfg xxx --model_path xxx

LICENSE

This repo is under the Apache-2.0 license. For commercial use, please contact the authors.

Owner
Sea AI Lab
Sea AI Lab
GitHub Repository
Rofi script to minimize / unminimize multiple windows in qtile

Qminimize Rofi script to minimize / unminimize multiple windows in qtile Additional requirements : EWMH module fuzzywuzzy module How to use it : - Clo

9 Sep 18, 2022
Improve current data preprocessing for FTM's WOB data to analyze Shell and Dutch Governmental contacts.

We're the hackathon leftovers, but we are Too Good To Go ;-). A repo by Lukas Schubotz and Raymon van Dinter. We aim to improve current data preprocessing for FTM's WOB data to analyze Shell and Dutc

ASReview hackathon for Follow the Money 5 Dec 09, 2021
A collection of daily usage utility scripts in python. Helps in automation of day to day repetitive tasks.

Kush's Utils Tool is my personal collection of scripts which is used to automated daily tasks. It is a evergrowing collection of scripts and will continue to evolve till the day I program. This is al

Kushagra 10 Jan 16, 2022
Python / C++ based particle reaction-diffusion simulator

ReaDDy (Reaction Diffusion Dynamics) is an open source particle based reaction-diffusion simulator that can be configured and run via Python. Currentl

ReaDDy 46 Dec 09, 2022
Checkers Project Built Using Python

Checkers Project Built Using Python

Meekness Anyaeche 1 Nov 08, 2021
Roblox Limited Sniper For Python

Info this is version 2.1 version 3 will support more options (install python: https://www.python.org) the program will buy any limited item with a pri

1 Dec 09, 2021
Data wrangling & common calculations for results from qMem measurement software

qMem Datawrangler This script processes output of qMem measurement software into an Origin ® compatible *.csv files and matplotlib graphs to quickly v

Julian 1 Nov 30, 2021
A collection of existing KGQA datasets in the form of the huggingface datasets library, aiming to provide an easy-to-use access to them.

KGQA Datasets Brief Introduction This repository is a collection of existing KGQA datasets in the form of the huggingface datasets library, aiming to

Semantic Systems research group 21 Jan 06, 2023
CuraMultiplyByGrid - Cura Плагин для размножения детали сеткой на весь стол автоматически без поворота

CuraMultiplyByGrid Cura Плагин для размножения детали сеткой на весь стол автоматически без поворота. Размножение в куре настолько ужасно реализовано,

3 Dec 02, 2022
JLC2KICAD_lib is a python script that generate a component library for KiCad from the JLCPCB/easyEDA library.

JLC2KiCad_lib is a python script that generate a component library (schematic, footprint and 3D model) for KiCad from the JLCPCB/easyEDA library. This script requires Python 3.6 or higher.

Nicolas Toussaint 73 Dec 26, 2022
These are my solutions to Advent of Code problems.

Advent of Code These are my solutions to Advent of Code problems. If you want to join my leaderboard, the code is 540750-9589f56d. When I solve for sp

Sumner Evans 5 Dec 19, 2022
El Niño - Southern Oscillation analysis compared to minimum flow rates of rivers in northeast Brazil

ENSO (El Niño - Southern Oscillation) analysis in northeast Brazil É comprovada a influência dos fenômenos El Niño e La Niña nas secas no nordesde bra

Weyder Freire 1 Jan 13, 2022
A python script to make leaderboards using a CSV with the runners name, IDs and Flag Emojis

SrcLbMaker A python script to make speedrun.com global leaderboards. Installation You need python 3.6 or higher. First, go to the folder where you wan

2 Jul 25, 2022
Explores the python bytecode, provides some tools to access it for fun and profit.

Pyasmtools - looking at the python bytecode for fun and profit. The pyasmtools library is made up of two parts A python bytecode disassembler . See Py

Michael Moser 299 Jan 04, 2023
A responsive package for Buttons, DropMenus and Combinations

A responsive package for Buttons, DropMenus and Combinations, This module makes the process a lot easier !

Skr Phoenix YT 0 Jan 30, 2022
Library for managing git hooks

Autohooks Library for managing and writing git hooks in Python. Looking for automatic formatting or linting, e.g., with black and pylint, while creati

Greenbone 165 Dec 16, 2022
🏃 Python3 Solutions of All Problems in GKS 2022 (In Progress)

GoogleKickStart 2022 Python3 solutions of Google Kick Start 2022. Solution begins with * means it will get TLE in the largest data set. Total computat

kamyu 38 Dec 29, 2022
An advanced NFT Generator

NFT Generator An advanced NFT Generator Free software: GNU General Public License v3 Documentation: https://nft-generator.readthedocs.io. Features TOD

NFT Generator 5 Apr 21, 2022
Eros is an expiremental programming language built using simple Python code.

Eros is an expiremental programming language built using simple Python code. Featuring an easy syntax and unique features like type slicing, the language remains an expirement that grows in down time

zxro 2 Nov 21, 2021
Flight Reservation App With Python

Flight Reservation App With Python

victor-h. 1 Nov 21, 2021