Accurateinfosolution/Suspicious_activity_detection_Yolov11_Custom

Custom YOLO Object Detection Model For Suspicious Activity Detection

This repository contains a custom-trained YOLO-based object detection model designed to detect various objects related to safety and surveillance scenarios, such as Assault, Fighting, Gun, Kidnapping, and more.

---

Table of Contents

• Overview
• Features
• Setup
• Usage
• Model Details
• Results
• Limitations
• Acknowledgments

---

Overview

The custom YOLO model is capable of detecting the following classes:

1. Assault
2. Fighting
3. Gun
4. Kidnapping
5. Knife
6. People
7. Police
8. Prisoner
9. Theft/Robbery
10. Time Bomb

The model is optimized for real-time inference and is suitable for applications in:

• Public Safety Monitoring
• Surveillance Systems
• Security Applications

---

Features

• Real-Time Detection: Utilizes YOLO’s efficient architecture for quick detection.
• Custom Classes: Trained on a tailored dataset specific to safety and surveillance.
• Confidence Thresholding: Outputs only high-confidence predictions for accurate results.
• Visualizations: Overlay detected objects on video feeds or images.

---

Setup

Prerequisites

• Python 3.7+
• ultralytics package for YOLO
• A GPU (recommended for faster inference)

Installation

1. Clone this repository:

   git clone https://huggingface.co/Accurateinfosolution/Suspicious_activity_detection_Yolov11_Custom
   

2. Install dependencies:

   pip install ultralytics
   

3. Add your trained YOLO model:

   • Place your Suspicious_Activities_nano.pt file in the project directory.

---

Usage

Running Predictions

Run the detection script on a video or image source:

python predict.py

Prediction Script: predict.py

The script uses the YOLO framework to load the best.pt model and detect objects:

from ultralytics import YOLO

# Load the custom YOLO model
model = YOLO("Suspicious_Activities_nano.pt")

# Perform prediction
model.predict(source="0", show=True, conf=0.6)

Replace "0" with:

• A webcam source (0 or 1).
• A path to an image/video file.

---

Model Details

• Architecture: YOLO (You Only Look Once)
• Classes: 10 custom classes (see Overview)
• Confidence Threshold: 0.6
• Pre-trained Weights: Fine-tuned on YOLO’s default weights.

---

Results

Sample output of the model includes bounding boxes for detected objects with labels and confidence scores. Images or video streams are displayed with real-time overlays.

---

Limitations

• Dataset Dependency: The accuracy is influenced by the quality of the training dataset.
• Environmental Constraints: Performance may vary under poor lighting or occlusion.
• Class Generalization: May not generalize well to scenarios outside its training scope.

---

Acknowledgments

• YOLO Framework: Ultralytics
• Custom Dataset: Annotated with bounding boxes for the 10 custom classes.
• Contributors: Shreyanth(mailto:shreyanthhg1427@gmail.com)

---