This repository collects all works developed related to Machine Learning and Image Processing, focusing on deep learning models for image classification and face recognition.
Keras MNIST: Classification of handwritten digits using Convolutional Neural Networks (CNNs) based on LeNet architecture. CIFAR-10: Image classification using CNNs. CIFAR-10 (Data Augmented): Improved classification performance through data augmentation techniques. VGG16 [-3]: Experimentation with a pre-trained VGG16 model for classification. STL-10 (Binary Classification): Using pre-trained models, but methodology for prediction with a generator is still under refinement. Pre-trained Model Predictions: Evaluations using ResNet50, VGG16, VGG19, and InceptionV3. Face Recognition (FACENET): Implementing a face recognition system using the FACENET model.
To explore and implement deep learning techniques for image classification, object recognition, and face recognition, leveraging both custom architectures and pre-trained models.
Programming Language: Python Deep Learning Frameworks: TensorFlow, Keras Pre-trained Models: VGG16, VGG19, ResNet50, InceptionV3, FACENET Image Processing Libraries: OpenCV, PIL (Pillow), NumPy, Matplotlib Data Augmentation Techniques: Keras ImageDataGenerator
Used CNN architectures for image classification. Experimented with transfer learning by fine-tuning pre-trained models. Applied data augmentation to enhance model generalization. Developed a Face Recognition pipeline using FACENET, including feature extraction and distance-based similarity measures.
Successfully classified MNIST and CIFAR-10 datasets with high accuracy. Improved classification performance using pre-trained models and data augmentation techniques. Demonstrated the effectiveness of FACENET for face recognition tasks. Provided a foundation for further research in AI-driven visual applications, including object detection, facial analysis, and real-time image processing.