In this repository, I intend to implement a facial recognition-based Attendance System with more than one hundred types of individuals. I have tried and compared many ways, such as using the facial recognition library to integrate face characteristics and a deep learning strategy by employing PyTorch (ResNet34) to develop a model capable of handling facial recognition tasks for more than a hundred individuals.

Varied models have different advantages and disadvantages, with some being particularly user-friendly, quick to train, or embedded. However, if insufficient data is available for training, it may not provide accurate results, such as just 50% efficiency. You may wish to include more models for comparison.

To do this, we have utilized a variety of techniques and strategies to attain greater precision and speed than anticipated. Whoever is interested in this repository. Because I have not only trained the model in Github, you should possess understanding of different technologies.

I've worked in multiple locations to train and incorporate the model as a.pth or .pickle file before deploying it on my Mac, as it is not compatible with most deep learning algorithms. If I have not provided a link to the Python file. Mention me in any context.

Due to the use of my device (Macbook M1). I develop most of the code in different environments. You might find the trouble with my code. I would strongly suggest that if you develop on an apple chip, you might need to swap between your local device (to test face recognition with a Laptop camera) and Google Colab to develop a specific model.

You might need to install requirements.txt separately for different tasks in this repo but I will prefer normal requirements for doing the normal task.

• Requirements: Python ,OpenCV, Deep Learning, PyTorch, Tensorflow
• Programs: VSC, Colab
• Libraries: OpenCV, TensorFlow, PyTorch

• facial_recognition (lib): Due to the accuracy compared to another deep learning model, this model is not used. Pros include that it's quite simple to learn and easy to comprehend.

• ResNet34 (with PyTorch): is now in use. It provides tremendous precision. Slightly more difficult than facial recognition as it requires knowledge of deep learning and the PyTorch framework.

• FaceNet: Next is FaceNet - Already write CNN model for embedding - still on the try a lot of problem with environments.

• VGGFace:

• Analysis of different face detection and recognition models for Android

• A Comparative Analysis Of Face Recognition Models On Masked Faces

• A Comparative Study of Facial Recognition Techniques