Facial Keypoints Detection Built a system to detect 68 distinguished keypoints; Consists of Haar Cascade algorithm for detection of faces and textbf CNN architecture for keypoints prediction.
Trained on images from YouTube Faces DataBase by employing Transfer Learning on Resnet-18 model Used Pytorch and OpenCV. Files included in this repository are all .ipynb files.
1_preprocessing.ipynb In this notebook,I prerpocessed the images and transformed into modelling form with custome classes based on Pytorch's "DATALOADER","TRANSFORMS" class.
2-architecture and modellingipynb Created whole pipeline for transformed to dataset into modeeling input form then modelled with resnet.
3-final pipelineipynb takes any image as input,extracts the faces using Har Cascade classifier and the predicts keypoints on those faces.
4-evaulaation.ipynb
5-fun.ipynb and experiment.ipynb experiments and put glass on the image using predicted keypoimts.
Here is How the result on my own photos. Pretty decent!!
