You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
A Gradio-based web application that detects whether an image is a deepfake. The application uses a pre-trained InceptionResnetV1 model from the facenet_pytorch library for face recognition, and pytorch-grad-cam for visual explainability.
A discriminative few-shot learning approach for face recognition and verification using a Siamese network architecture. Employing a triplet loss function, the model optimizes the embedding space to cluster faces of the same individual and separate those of different individuals, enhancing accuracy and efficiency with limited training data.
Docker and Flask based API layer + data ingestion pipeline for the Facenet-PyTorch facial recognition library. I.e. simple ML deployment for matching pairs of photos
TubeFaceCrop is a tool for crawling videos related to keywords from YouTube and preprocessing them. It is based on MTCNN to automatically remove faces from the videos and perform central crop, and finally segments the videos into 5-second clips for further processing and analysis.
🎬 VideoDeepFakeDetection uses AI to authenticate videos through a multi-step process, identifying potential deepfakes for enhanced content reliability.
This project is an implementation of a secure and efficient voting system utilizing facial recognition technology. The system aims to enhance the integrity and convenience of the voting process by allowing voters to authenticate themselves through facial recognition before casting their votes.
Developed a working prototype of a University Security Monitoring System using Deep Learning to assist Reva University in maintaining a database of known and unknown persons entering through the entrance gate by detecting the frontal faces of students.
This repository hosts a cutting-edge facial recognition system designed to enhance customer identification and verification. Leveraging MTCNN for accurate face detection and DeepFace-FaceNet for facial embeddings, the system integrates with Pinecone's vector database to efficiently match and verify repeat customers.
This repository is home to an exploration in the field of Facial Recognition using Convolutional Neural Networks. It is based on the performance comparison between different models such as ResNet50, MobileNetV3, InceptionV3, EfficientNet, and VGG16. The models were trained using two types of losses - Triplet Loss and Categorical Cross Entropy.