futurice/intro-to-object-detection

Name: intro-to-object-detection

Owner: Futurice

Description: instruction and sample codes to get started with object detection in dlib-python

Created: 2017-08-14 09:01:58.0

Updated: 2018-05-03 19:41:41.0

Pushed: 2017-09-01 08:51:09.0

Homepage: null

Size: 9916

Language: Python

GitHub Committers

User	Most Recent Commit	# Commits

Other Committers

User	Email	Most Recent Commit	# Commits

README

Introduction to Object Detection

This is a starting template to learn how to build an object detector. We want to detect images of a specific object from digital photographs. In order to achieve this goal, we shall employ the help from opencv3, a computer vision library, for image manipulation, and dlib, a machine learning library.

Notes

• Slides at link
• The recommended programming language is Python. However, it is possible to adapt to C++ easily.
• It is recommended to use a Python virtual environment. Instal virtualenv with pip.
• The dlib version from pip does not include all dlib features. It is recommended to build dlib from source.

Setup environment

1. Install boost-python

   X
    install boost-python --with-python3
   

Ubuntu

sudo apt-get install libboost-all-dev

reate virtual python environment (optional)

virtualenv .env -p python3.6 source .env/bin/activate

nstall dependencies (`opencv3` and `dlib`)

pip install -r requirements.txt

leanup (optional, in case you want to recreate your python environment)

deactivate rm -rf .env

Challenge instruction
are to build an image object detector (that detects object images in photograph). In order for your detector to work, you need to teach your detector how to recognize a certain object (e.g. a car, a house, a dog, a cat, ...). The steps to teach your detector are as follow:
ind an image source of photographs of your object (Google, Bing, DuckduckGo).
anually label your object in the photographs with `imglab` (provided).
reate an SVM from the labeled data (this is when you start coding).
se the trained SVM as your object detector.

# How to use `imglab`
gister images to label

imglab -c training.xml imglab -c testing.xml

bel images

imglab training.xml