oarriaga/face_classification
Face Classification
Este proyecto es un sistema de clasificación facial mediante deep learning que detecta rostros humanos y clasifica género y emoción. Utiliza redes neuronales convolucionales y herramientas de visión artificial para analizar atributos faciales tanto en imágenes estáticas como en transmisiones de video en vivo.
El sistema incluye clasificadores especializados para emociones basados en el dataset FER2013 y para género basados en datasets de IMDB. Estos modelos están integrados en un servicio web contenedorizado, permitiendo que la lógica de clasificación se exponga como una API que procesa datos de imagen mediante peticiones de red.
La superficie técnica cubre todo el pipeline, desde el entrenamiento de modelos personalizados con Keras hasta la detección facial en tiempo real usando OpenCV. Esto incluye la implementación de un pipeline de video para el procesamiento continuo de fotogramas y el uso de Haar cascades para localizar rostros.
Features
- Emotion Classifiers - Classifies human emotions and gender from images and live video using deep learning networks.
- Face Detection - Identifies human faces within live video streams to prepare data for downstream attribute analysis.
- Custom Model Training - Implements custom training pipelines using Keras to create gender and emotion classification models from labeled image datasets.
- Attribute Classifiers - Combines face detection with Keras-based CNNs to classify gender and emotion.
- Facial Emotion Classifiers - Provides a vision tool that identifies human emotions based on the FER2013 dataset.
- Facial Expression Recognition - Analyzes facial muscle patterns to categorize emotional states and gender using trained neural networks.
- Gender Predictions - Implements a neural network trained on IMDB datasets to predict the gender of subjects in photographs.
- CNN Architectures - Implements convolutional neural network architectures to extract visual patterns for gender and emotion classification.
- Computer Vision - Implements computer vision capabilities for facial recognition and classification deployed as a web service.
- Keras Model Implementations - Utilizes the Keras API to implement and train deep learning architectures for facial attribute classification.
- Cascade Classifier Detections - Uses OpenCV Haar cascade classifiers to locate human faces by scanning for edge and line patterns.
- Real-Time Facial Recognition - Performs low-latency identification of emotions and gender from live video streams.
- Face Recognition and Analysis - Detects human faces and analyzes associated attributes like gender and emotion within image frames.
- Real-Time Model Inference on Frames - Implements a high-speed video pipeline that runs model inference on every frame of a live camera stream.
- Model Training Implementations - Implements the training of custom classification models for gender and emotion using Keras and labeled datasets.
- Image Classification APIs - Ships a RESTful API that accepts image data and returns classification results for facial attributes.
- Containerized Service Deployments - Packages the classification service within containers to ensure consistent deployment and inference across environments.
- Inference APIs - Exposes the facial analysis model as a web service that accepts image data and returns classification results.
- Age gender estimation - Listed in the “Age gender estimation” section of the Ailia Models awesome list.
