Server Start Guide - person-in-hangang/HanRiver GitHub Wiki
Introduction
Server consists of official PyTorch implementation of [1] Camera Distance-aware Top-down Approach for 3D Multi-person Pose Estimation from a Single RGB Image (ICCV 2019) and [2] Pedestrain Attribute Recognition.
-
Camera Distance-aware Top-down Approach for 3D Multi-person Pose Estimation from a Single RGB Image (ICCV 2019) contains [3]RootNet part and [4]PoseNet part.
-
RootNet & PoseNet : These are used to estimate person's height & fallen detection
-
Pedestrain Attribute Recognition(PAR) : This is used to estimate person's three most probable features contains [2] Pedestrain Attribute Recognition.
Dependencies
Python 3.6 version with PyCharm is used for development
Start Guide For Socket Programming
-
Install Python and Python development tool (ex. PyCharm) on your computer
-
Download project
-
Download pre-trained RootNet in here, and pre-trained PoseNet in here, then move them to demo folder
-
Go to mysocket.py in demo folder
-
Set your port number and path to save and load images
-
Run mysocket.py
-
Connect 119 App, Detection camera and Tracking camera then operate them
-
You can see images received from connected Detection camera and Tracking camera
-
Then Server will send calculated information to connected 119 App
More details
This project offer pre-trained RootNet and PoseNet model.
If you want to newly train model, Please refer RootNet part and PoseNet part.
Start Guide For pedestrian attribute recognition
- Download and prepare the dataset and pretrained model as follow:
PA100K Links
./dataset/pa100k/annotation.mat
./dataset/pa100k_ckpt_max.pth
Reference
[1] : Camera Distance-aware Top-down Approach for 3D Multi-person Pose Estimation from a Single RGB Image (ICCV 2019), https://arxiv.org/abs/1907.11346
[2] : Pedestrain Attribute Recognition, https://github.com/valencebond/Strong_Baseline_of_Pedestrian_Attribute_Recognition13
[3] : RootNet, https://github.com/mks0601/3DMPPE_ROOTNET_RELEASE
[4] : PoseNet, https://github.com/mks0601/3DMPPE_POSENET_RELEASE