🚀 Fall-Detection: SVM-based Fall Event Detection

Detecting fall and not-fall events in videos using SVM

This repository provides a simple pipeline for detecting fall events in video sequences using a Support Vector Machine (SVM) classifier.

---

📌 Table of Contents

1. Overview
2. Data & Feature Extraction
3. Training
4. Testing / Detection
5. Results
6. Acknowledgements

---

🧠 Overview

This project detects fall vs. not-fall events in videos:

• Features are extracted from video frames.
• A Support Vector Machine (SVM) is trained on these features.
• The trained SVM is used to classify new video sequences.

---

📂 Data & Feature Extraction

1. Define the path to save extracted features of fall and not-fall events from training videos.
2. Define the path to the training videos.
3. Define paths to all extracted features (new and old).
4. To extract features for fall and not-fall events during runtime, press the z key.

⚡ Feature extraction is required before training the SVM.

---

🛠 Training

1. Define the path where the trained SVM should be saved.
2. Run the training script with your extracted features to train the classifier.

The SVM model will learn to distinguish between fall and not-fall events based on the features.

---

🖥 Testing / Detection

1. Define the path to the trained SVM.
2. Define the path to a test video.
3. Run the detection script to classify fall and not-fall events in the video.

The SVM predicts frame-wise or event-wise labels using the learned model.

---

📊 Results

Below is an example output of the fall-detection system:

result.mp4

• The dataset for training and testing was collected by Mr. Farahnezhad, who also contributed to code development.
• This video shows a demonstration of fall vs. not-fall detection in action.

---

🙏 Acknowledgements

We thank Mr. Farahnezhad for providing the dataset and supporting the development of this project.