This project focuses on building a robust system for censoring explicit and harmful content within images and videos. We achieve this by fine-tuning a YOLO (You Only Look Once) model, a cutting-edge object detection algorithm, to accurately identify and pinpoint such content. The model is meticulously trained on diverse datasets to recognize a wide range of problematic elements, ensuring comprehensive coverage and high detection accuracy.

Once the YOLO model identifies instances of explicit or harmful content, it provides precise bounding box coordinates for each detected item. This critical information is then seamlessly passed to a straightforward Python script. This script leverages the provided coordinates to automatically apply blurred boxes over the identified regions, effectively obscuring the problematic content while preserving the surrounding visual information. This automated approach offers a swift and efficient solution for content moderation, significantly reducing the need for manual review.

To get this project up and running, follow these simple steps. Using a virtual environment is highly recommended to keep your project dependencies organized and avoid conflicts with other Python projects.

Open your terminal or command prompt and run the appropriate command below to create a virtual environment named venv.

For Windows:

python -m venv venv

For macOS and Linux:

python3 -m venv venv

You'll need to activate the virtual environment to ensure all installed packages are isolated to this project. Choose the command that applies to your operating system:

For Windows:

.\venv\Scripts\activate

For macOS and Linux:

source venv/bin/activate

Once your virtual environment is active, install the necessary libraries. The easiest way is to use the requirements.txt file (if provided). Use the command appropriate for your system:

For Windows:

pip install -r requirements.txt

For macOS and Linux:

pip3 install -r requirements.txt

If you don't have a requirements.txt file or prefer to install them individually, you can use the following commands instead:

For Windows:

pip install opencv-python-headless pandas ultralytics numpy

For macOS and Linux:

pip3 install opencv-python-headless pandas ultralytics numpy

Create a store directory in the root folder cause it will contain following necessary folders:

• custom_models: It contains newly trained YOLO models after training.
• dataset: It contains all the datasets that are used in training.
• results: It contains following files:
  • detection_results_insights.json: It contains the insights of detection_results.json, like which object class appears how many times.
  • detection_results.json: It contains the information and annotation of detected objects in a video frame-by-frame.
  • predictions.json: File containing model prediction results.
• videos: This folder contains the videos to be user in object detection, and a folder containing censored videos.
  • censored: It contains only videos that are being censored based on the information given by model in detection_results.json.

1. Download After setting up virtual environment and installing necessary dependencies, download the dataset from here.
2. After downloading unzip the dataset in the dataset folder.

1. Make sure you have the videos and dataset setup in the right folders as documented above.

2. Check all the paths and variables in the main.py script are as per your folder structure.

3. Proceed with running the script

For Windows:

python main.py

For macOS and Linux:

python3 main.py

[Citation details will be added after publication of the paper]

This project is licensed under the MIT License. Please refer to the LICENSE for details.

For any questions, suggestions, or issues, please open an issue in this repository or contact mdazlaan1996@gmail.com.