Skip to content

Real-time selfie filters using facial keypoints regression and opencv

Notifications You must be signed in to change notification settings

agrawal-rohit/opencv-facial-filters

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

12 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

OpenCV Facial Filters

bleh

Overview

This project demonstrates the creation of Snapchat-like facial filters using Deep Learning and OpenCV. It involves facial keypoint detection to superimpose themed filters on a face in real time. I came up with the idea for the project on 21 December 2018, Hence the Christmas theme :P.

Detailed methodology and insights can be found in this Medium article.

Dataset

Utilized the Facial Keypoints Detection dataset from Kaggle.

Methodology

  • Data Augmentation: Flipped images and key points for diversity.
  • Model Architecture:
    • CNN: Acts as a feature extractor.
    • ANN: A fully connected network for facial keypoint regression.
  • Training:
    • Loss Metric: Mean Absolute Loss.
    • Performance: Achieved ~0.0113 after 300 epochs with the Adam optimizer.
  • Implementation:
    • Real-time Data Capture: Used OpenCV for live webcam feed.
    • Preprocessing: Standardized input before feeding into the model.
    • Output Utilization: Keypoint positions determined the placement and scale of thematic filters.

About

Real-time selfie filters using facial keypoints regression and opencv

Topics

Resources

Stars

Watchers

Forks

Sponsor this project

 

Languages