Guardian Bot

• 

  Output Depicting the categorization of video as not violent

• 

  Output Depicting the categorization of video as violent

Inspiration:

The inspiration behind Guardian Bot is to create a real-time violence alarm system utilizing CCTV surveillance. The goal is to automate the identification of violence in CCTV videos, which would be impractical to monitor manually. The system aims to provide instant alerts when violence is detected, potentially saving lives and promoting community safety.

What it does:

Guardian Bot is a real-time violence alarm system that uses CCTV surveillance to detect instances of violence in videos. It analyzes the CCTV footage and alerts people immediately when violence is identified, allowing them to take protective measures such as contacting the authorities, hiding, or fleeing.

How we built it:

The system is built using the ResNet50 model, which is employed to extract features from CCTV videos. These features are then used to train a classifier to distinguish between violent and non-violent videos. The process involves collecting a dataset of CCTV videos containing both violent and non-violent situations, preprocessing the videos by resizing and converting them to grayscale, and then utilizing the ResNet50 model for feature extraction.

Challenges we ran into:

The team faced various challenges during the development of Guardian Bot. Some of these challenges include collecting a diverse and representative dataset of CCTV videos, optimizing the performance of the ResNet50 model for real-time processing, and ensuring the system's accuracy and reliability in identifying violence accurately.

Accomplishments that we're proud of:

The team is proud to have created a functional real-time violence alarm system using CCTV surveillance. The system's potential to save lives, prevent violence, and enhance community safety is a significant accomplishment. Additionally, successfully implementing the ResNet50 model for violence detection demonstrates the team's technical prowess.

What we learned:

During the development process, the team gained valuable insights into using deep learning models like ResNet50 for violence detection in CCTV videos. They also learned about the challenges associated with real-time processing, dataset collection, and system reliability.

What's next for Guardian Bot:

In the future, Guardian Bot aims to incorporate multimodal sensing, integrating various sensing modalities like audio analysis and body language recognition. This will provide a more comprehensive understanding of potentially violent situations, leading to more accurate and timely alerts. The system also plans to be easily scalable and adaptable to different environments, integrate geolocation data for tracking violence patterns, and collaborate with law enforcement and emergency response services for a quicker and more effective response to incidents.

Built With

• imgaug
• kaggle
• keras
• machine-learning
• os
• pipeline
• resnet
• tensorflow