pallabganguly / opcv-project

Detection of Hands using HAAR Cascades in OpenCV for UG Project

Detection of Open Hand using HAAR Cascades

opcv-project

Detection of Hands using HAAR Cascades in OpenCV for UG Project Souce code, classifiers and associated data for UnderGraduate Project. This repository deals with detection of open hand using OpenCV 3.1.0 on Python 3.5. Data files are omitted for now

Requirements:

• OpenCV for Python 3.5
• numpy
• matplotlib

Work done:

In this repository, we explore the effects of various parameters on accuracy of detection. For training, about 8000 samples were used, and for testing, about 3000. We then tweak:

1. Size of window
2. No of neighbours
3. Image scaling We run the classifier after changing the above parameters one at a time, on a test set to generate plots of accuracy versus the parameter in question. If you want the data for re-training, please contact [email protected]

To run:

There are scripts to generate plots of accuracy versus several parameters of HAAR classifier Navigate to ./src and run python .py to run the respective script

Results:

Note: This repository contains original research work done by the author. Please include proper citations when using any of the results obtained from this repository.

opcv-project's Issues

Paths for barplot are also incorrect

Needs to be prefixed correctly classifiers = ['cascade_30x30.xml', 'cascade_40x40.xml', 'cascade_50_10.xml', 'cascade_50_15.xml']

Paths for loading classifier needs to be corrected

In the files src/newsample.py, letsrunit.py, paths are not correct
In newsample hand_cascade = cv2.CascadeClassifier('cascade_3050_10.xml')