This project aims to contribute to surveillance technology by introducing an automated person recognition system. Leveraging deep learning, specifically Faster R-CNN and ResNet50, the system enhances the identification and localization of individuals in CCTV footage, addressing the inefficiencies of manual monitoring.
git clone <repository-url>
cd Smart-Surveillance
cd "Attribute recognition"
pip install -r requirements.txtUsage To start the person recognition process, execute:
python person_recognition.py --input <path_to_video>- Automated Person Detection: Utilizes Faster R-CNN for real-time object detection in video footage.
- Attribute Recognition: Employs fine-tuned ResNet50 on the PETA dataset for precise attribute matching, including clothing and accessories.
- Query-Based Person Mapping: Processes natural language queries to match textual descriptions with visual attributes, enhancing the search and identification process.
The project uses two main datasets:
- UCF Crime Dataset: For analyzing real-world surveillance videos.
- PETA Dataset: Known for its detailed pedestrian attribute annotations, aiding in attribute recognition.
Please go through the pdf document for further information regarding the project including the results and accuracies.
Distributed under the MIT License. See LICENSE for more information.
This project builds upon existing research in the fields of computer vision and machine learning, aiming to provide a practical solution for law enforcement and public safety.