Comparison between chainer and pytorch about implementing DeepPose

This repository contains codes for Qiita blog post.

Usage

Dataset preparation

First download and prepare datasets for training by running the following script.

python scripts/dataset.py

Start training

If you want to train model with Chainer, just run:

python scripts/train.py chainer

In the case of PyTorch:

python scripts/train.py pytorch

If you want to run train.py with your own settings, please check the options by python scripts/train.py --help and give customized training settings.

Visualize

Visualizing training time, just run:

python scripts/plot_training_time.py "title of graph"

Visualizing inference time, just run:

python scripts/plot_estimating_time.py 10000 "title of graph"

Install

pip install pyflakes pycodestyle
pip install sphinx
pip install -r requirements.txt