Actors Face Recognition

This repository contains the code to replicate the wave app, which recognizes the faces of the actors Aaron Eckhar, Adam Brody, Bradley Cooper, and Adrien Brody

1. Clone this repo

git clone https://github.com/ksathur/Wave-App--Actors-Face-Recognition.git
cd Wave-App--Actors-Face-Recognition

2. Environment

This code runs on Python 3.7

2.1 Create a conda environment with python 3.7 and activate it

conda create -n face_recog_wave python=3.7
conda activate face_recog_wave

2.2 Pip install requirements.txt

This will install all the libraries that are require to run the wave app testing.

pip install -r requirements.txt

3. Run

wave run wave_face_recognition.py

Go to http://localhost:10101/actors_face_detection to visualize the app. Screenshot of the interfaces are available at ./interface_1.PNG, and ./interface_2.PNG. A recorded demo video is available at ./demo_video.mp4.

4. Custom Training

4.1 Upload the folder contains the images of each person into ./dataset/raw.

4.2 Change the directory to ./utils.

cd utils

4.3 Run face_csv_file_writter.py to create a .csv file which consists of images directories and labels.

python `face_csv_file_writter.py`

4.4 Go back to previous working directory.

cd ../

4.5 Run face_classifier.py to start the training. The model and the accuracy information will be saved into ./dumps/model and ./dumps/accuracy respectively.

python `face_classifier.py`

4.6 Run face_tester.py to evaluate the peformance of the model on validation dataset.

python face_tester.py

4.7 Run plot.py to visualize the performance (via confusion matrix) (./dumps/accuracy/cfm.png).

python plot.py

4.8 To run testing on any test data download the images into ./test_sample and run the face_classifier_sample_test.py.

python face_classifier_sample_test.py