Related Experiment Video
Updated: Jul 6, 2025
13:19
Deep Neural Networks for Image-Based Dietary Assessment
Published on: March 13, 2021
9.2K
Automated facial characterization and image retrieval by convolutional neural networks
Syed Taimoor Hussain Shah1, Syed Adil Hussain Shah1,2, Shahzad Ahmad Qureshi3
1PolitoBIOMed Lab, Department of Mechanical and Aerospace Engineering, Politecnico di Torino, Turin, Italy.
Frontiers in Artificial Intelligence
|January 4, 2024
Summary
This study introduces FRetrAIval (FRAI), a novel hybrid deep learning model for facial feature extraction and identification. FRAI achieves high accuracy in facial recognition tasks, showing potential for applications in healthcare and criminology.
Area of Science:
- Computer Vision
- Machine Learning
- Biometrics
Background:
- Inferring relationships between facial images, especially those with varying expressions or over time, remains a challenge in imaging research.
- Accurate facial feature extraction and identification are crucial for various applications, including medical diagnosis and forensic science.
Purpose of the Study:
- To develop and evaluate a novel method for facial feature extraction, characterization, and identification.
- To present a hybrid deep learning system that combines classical computer vision with convolutional neural networks.
Main Methods:
- A hybrid deep learning model, FRetrAIval (FRAI), was developed, integrating GoogleNet and AlexNet architectures.
- Computer vision techniques, including an oriented gradient-based algorithm, were used for face region extraction and preprocessing.
- The FRAI model was trained and validated using the Aligned Face dataset (AFD) and Labeled Faces in the Wild (LFW) dataset.
Main Results:
- The FRAI system demonstrated superior performance compared to existing methods, achieving high precision, recall, F1, and F2 scores on both AFD (up to 92.52%) and LFW (95.00%) datasets.
- The model effectively extracts and identifies facial features, outperforming previous techniques.
Conclusions:
- The developed FRAI model offers an efficient and accurate solution for facial feature extraction and identification.
- This technology holds significant potential for applications in healthcare, criminology, and other fields requiring rapid and reliable facial identification.
