xiehaizheng / Convolutional-Pose-Machines-Pytorch

Pytroch version of Convolutional Pose Machines

Convolutional-Pose-Machines-Pytorch

This is a pytroch version of Convolutional Pose Machines, you can find more details in the original paper Convolutional Pose Machines

Results

The result is generated by the model, which has trained 100 epoches.

Requirements

• Python 2.7
• PyTorch 0.3.0
• torchvision 0.2.0
• OpenCv > 2.4.9

The data set

• Download the data set for training LSPet
• (Alternative) Download the data set for validation LSP
• When you want to train some other datasets, please change the code to correspond to your datasets.

Usage

Train

• Set the train parameters in " config/config.yml ".
• Go to the "train_val" folder cd train_val.

• with val

python cpm_train.py --gpu 0 1 --train_dir ../dataset/train/lspet_dataset --val_dir ../dataset/val/lsp_dataset --config ../config/config.yml

• without val

python cpm_train.py --gpu 0 1 --train_dir ../dataset/train/lspet_dataset --config ../config/config.yml

• You can also train the model like

sudo sh cpm_train.sh

Test

• Go to the "test" folder cd test.

python cpm_test.py

Pre-trained model

• cpm_checkpoint_pth.tar--Baiduyun Password: huaj

Notice

• The code of test is just used for visualization,you can change the code to test your own image on condition that you have trained another model "person-center-network"(similar to CPMs).
• I hope that'll be helpful for you.