Webcam Surveillance

This project detects humans in a live webcam feed using a MobileNet-SSD model. It saves screenshots when a person is detected and logs the duration of their presence, also starts video recording if the person stays for too long. The script can be run as a command-line tool on Linux or as a Python script on other platforms.

---

Features

• Real-time human detection using a pre-trained MobileNet-SSD model.
• Saves screenshots when a person is detected.
• Logs detection durations in a CSV file.
• Starts recording if the intruder stays for too long.
• Customizable configuration via config.json.

---

Installation

1. Clone the Repository

git clone https://github.com/FarhanSadaf/webcam.git
cd webcam

2. Set Up a Virtual Environment

python3 -m venv .venv
source .venv/bin/activate

3. Install Dependencies

pip install -r requirements.txt

---

Usage

On Linux (Using webcam Command)

Make the Script Executable:

chmod +x webcam

Run the Script:

• Without showing the video feed:

  ./webcam

• Video Recording Disabled:

  ./webcam --novideo

• With the video feed:

  ./webcam --show

• Print detection statistics:

  ./webcam --stat

• Print detection statistics for the last N days (e.g., last 2 days):

  ./webcam --stat 2

• Print top M results for the last N days (e.g., top 2 results for the last 3 days):

  ./webcam --stat 3 -top 2

• Delete output directories for specific days:

  ./webcam --delete

Add to PATH (Optional):

To use webcam as a global command:

sudo mv webcam /usr/local/bin/

On Non-Linux Systems

Run the Python Script Directly:

• Without showing the video feed:

  python main.py

  • Video Recording Disabled:

  python main.py --novideo

• With the video feed:

  python main.py --show

• Print detection statistics:

  python main.py --stat

• Print detection statistics for the last N days (e.g., last 2 days):

  python main.py --stat 2

• Print top M results for the last N days (e.g., top 2 results for the last 3 days):

  python main.py --stat 3 -top 2

• Delete output directories for specific days:

  python main.py --delete

---

Configuration

Modify config.json to customize the behavior:

• model_weights: Path to the model weights file.
• model_config: Path to the model configuration file.
• frame_rate: Desired frame rate.
• input_size: Input size for the model.
• output_folder: Folder to save screenshots and logs.
• camera_id: Change video capture camera.
• second_screenshot_interval: Time (sec.) after second screenshot will be taken or video recording will be started.
• recording_duration: Minimum amount of time (sec.) the video will be recorded.
• coverage_threshold: Average brightness threshold for dark frames. Adjust it according to your environment.

---

Example Output

Screenshots

Screenshots are saved in the output/<date>/ folder with timestamps.

CSV Log

Detection durations are logged in output/<date>/detection_durations.csv.

Statistics

Use --stat to print the CSV log in a table:

+---------------------------+-------------------+
|          Filename         | Duration (seconds)|
+---------------------------+-------------------+
| 2023-10-01_12-34-56_#1.jpg|       3.45        |
| 2023-10-01_12-35-10_#2.jpg|       7.89        |
+---------------------------+-------------------+

---

License

This project is licensed under the MIT License. See LICENSE for details.

---

Contributing

Feel free to open issues or submit pull requests. Contributions are welcome!