Skip to content

Latest commit

 

History

History
103 lines (69 loc) · 2.74 KB

README.md

File metadata and controls

103 lines (69 loc) · 2.74 KB
Raw

Full tutorial to train Haar Cascade using opencv3 on MacOS

Material:

  • MacOS(Over Catalina10.15)
  • CMake
  • Homebrew
  • Python3
  • Datasets-Over 1000 positive images and negative images

1.Install Supports(Homebrew,Cmake,OpenCV)

Install HomeBrew

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

Install Cmake

brew install cmake git subversion wget echo "export PATH=/usr/local/bin:$PATH" >> ~/.bashrc source ~/.bashrc

Install Python3

brew install python3

Install Opencv3

brew install opencv@3

2.Get ready datasets

For the training you need over 1000 positive images and negative images, positive images are images that only contains the item you want to detect and the negative images could be any image. You Could download these images from the following website just simpily search it up:

  1. www.kaggle.com
  2. https://datasetsearch.research.google.com

Then make a training folder and make another 2 folder named pos and neg in it.

mkdir train

cd train

mkdir pos

mkdir neg

Put all the positive images in to pos and negatives into neg

Reminder:Make sure the size of the positives are all equal and recommanded sizes are 20x20,45x45,100x100

3.Create Samples

We need some sample for the training business and we will be using the opencv_createsample program to do it.

First,we have to get the path where you opencv3 is installed at so run the following command:

brew --prefix opencv@3

The output should be like:

(base) mac@mac train % brew --prefix opencv@3

/usr/local/opt/opencv@3

Make sure you are in the training folder to generate these codes.

This generate the path for the positives

ls ./neg/.>neg.txt

This generate the path for the negatives

ls pos/. > pos.info

Open pos.txt and it should be like:

./pos/apple_1.jpg

./pos/apple_10.jpg

./pos/apple_100.jpg

./pos/apple_1000.jpg

Go to tools and replace, search jpg up and replace them with : jpg 1 0 0 yourheight yourwidth

Now Use this template:

YourPathToOpenCVopencv_createsamples -bg neg.txt -info pos.info -num 10000 -Your Width -Your Height -vec pos.vec

###4.opencv_traincascade

You should be result with a .vec files now then fill in this template to begin cascade training.

mkdir data

Your Path to opencvopencv_traincascade -data data -vec positives.vec -bg neg.txt -numPos 1000 -numNeg 1000 -numStages 8 -your width -your height -minhitrate 0.99 -maxFalseAlarmRate 0.3 -precalcValBufSize 1024 -precalcIdxBufSize 1024 -numThreads 6

This Would get you training and after it finish it should give you a .xml file in your data folder.

If you have any question or concerns you could go to the template Directory to look at what am I doing