Public | Automated Build

Last pushed: 18 days ago
Short Description
Human pose estimation implementation using Caffe, OpenCV 3.1.0 and Python 2.7.
Full Description

Video Pose Extractor

"Nature is a language can't you read?" - The Smiths

Dockerized, minimal implementation of Cao et al.'s repository including Caffe, OpenCV and Python 2.7.

Includes:

  • CPU Caffe implementation
  • python bindings + OpenCV for happy hacking
  • easy-to-use video-to-images script for pose finding in videos
  • dump limb data into a file for later analysis and modeling

Demo

Getting started

Install Docker

From the command line:

docker pull justinshenk/pose-extractor

or build it locally

git clone https://github.com/justinshenk/video-pose-extractor.git
cd video-pose-extractor
docker build -t justinshenk/pose-extractor:cpu -f .

Image

python convert.py -i [image]

Video

python convert.py -i [input_file] -s [start_time] -t [duration] -r [frame_rate]

Note: Time is in hh:mm:ss format.

Data is saved in positions.npy and can be loaded using

ipython
import numpy as np
raw = np.load('positions.npy')
data = raw.take(0)
Docker Pull Command
Owner
justinshenk
Source Repository

Comments (0)