Skip to content

We detect and track faces in video, then extract features from those face tracks and try to cluster them into given number of persons.

Notifications You must be signed in to change notification settings

ankuPRK/Face-Detection-Tracking-and-Clustering

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

18 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Face-Detection-Tracking-and-Clustering

We detect and track faces in video, then extract features from those face tracks and try to cluster them into given number of Clusters, each Cluster representing a unique person. To get a full idea of our project, refer to the file "ppt_explaining_the_project.pdf". Results were obtained on video: https://youtu.be/A1fVcj29xhk

Running the project on a video-file involves 4 steps:

  1. Shot Detection:

We need the information of shots from the video. The shot-detection is done using Python.

Application used is ShotDetect by John Mathe: https://github.com/johmathe/Shotdetect ReadMe on this GitHub page describes how to use the app via command line.

This app is available for Ubuntu: sudo apt-get install ShotDetect

After running app on your video, a folder will be generated. Paste the "xml_parser.py" file (available here) in that folder and run it.

A file by name "shot_info.txt" will be generated. This file contains shot information in a way understandable by our Face Tracking code. Copy this file and paste it at the location of "main_for_shots.cpp"

  1. Detection and Tracking:

You need to build an OpenCV project, with Dlib Library for Facial Landmark Points Detection, available at: https://sourceforge.net/projects/dclib/files/dlib/v18.10/dlib-18.10.tar.bz2/download

We need the DLib Detector to return the Landmark Points as vector of Points. For this, replace the "full_object_detection.h" file of Dlib Library with the one given here.

Now add the "main_for_shots.cpp" file in the OpenCV project with DLib. Add your video and its "shot_info.txt" generated in step-1 at the cpp file's location.

  1. Feature Extraction:

Copy the "data" folder created in Step-2, and paste it at the location of the Feature Extraction cpp files. Run the .cpp files then.

  1. Clustering:

Is done in Python. Copy the folders generated in step3 and paste them where the Python files are.

If features are generated using "avg_img_features.cpp", run "Clustering_basic_feats.py" results will be stored in "resultsAVG/".

If features are generated using "3D_track_features.cpp", run "Clustering_adv_feats.py" results will be stored in "results/".

About

We detect and track faces in video, then extract features from those face tracks and try to cluster them into given number of persons.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published