The Goal of the project is to do Face recognition task using Siamese Neural Network that is trained over a triplet loss function , This is useful in situations where you have limited dataset for training the model.
This codebase implements a face recognition system using a Siamese Neural Network architecture. It helps you understand how this approach works and tackles the challenge of large datasets by employing one-shot learning techniques.
Model Functionality:
Feature Extraction: Generates a 128-dimensional vector embedding for any input image, capturing its key features.
Similarity Measurement: Compares the 128-dimensional vectors of two images to determine their similarity and classify them as belonging to the same person or not.
Triplet Loss for Training:
The model leverages a triplet loss function during training. This function uses triplets of images: anchor (reference), positive (same person), and negative (different person). It aims to:
Minimize the distance between the anchor and positive embeddings.
Maximize the distance between the anchor and negative embeddings.
This approach helps the model learn robust representations for effective face discrimination.
The Goal of the project is to do Face recognition task using Siamese Neural Network that is trained over a triplet loss function , This is useful in situations where you have limited dataset for training the model.
This codebase implements a face recognition system using a Siamese Neural Network architecture. It helps you understand how this approach works and tackles the challenge of large datasets by employing one-shot learning techniques.
Model Functionality:
Feature Extraction: Generates a 128-dimensional vector embedding for any input image, capturing its key features.
Similarity Measurement: Compares the 128-dimensional vectors of two images to determine their similarity and classify them as belonging to the same person or not.
Triplet Loss for Training:
The model leverages a triplet loss function during training. This function uses triplets of images: anchor (reference), positive (same person), and negative (different person). It aims to:
Minimize the distance between the anchor and positive embeddings.
Maximize the distance between the anchor and negative embeddings.
This approach helps the model learn robust representations for effective face discrimination.