INF573 - Make Benard Disappear

This project applies advanced computer vision techniques to remove a specific individual wearing a red cap, whom we will refer to as Benard, from visual media such as photos, videos, and live recordings. The modular pipeline integrates state-of-the-art algorithms for detection, segmentation, tracking, and inpainting.

Table of Contents

• Overview
• Pipeline Workflow
• Installation
• Usage
• License

Overview

This project explores various applications of computer vision, such as:

• Detection: Identifying the individual using YOLO and a red cap heuristic.
• Segmentation: Creating pixel-level masks of the individual for further processing.
• Tracking: Ensuring continuous detection across video frames.
• Inpainting: Reconstructing the background after removing the individual.

Pipeline Workflow

The pipeline consists of the following steps:

1. Detection:

   • YOLO identifies all individuals in the frame.
   • A heuristic-based red cap detection isolates areas potentially containing a red cap.
   • Results are combined to locate the individual wearing the red cap.

2. Segmentation:

   • Generates pixel-level masks for the detected individual.

3. Tracking:

   • KCF/MOSSE trackers ensure continuity in detection during temporary detection failures.

4. Inpainting:

   • Removes the identified individual and fills the void with reconstructed backgrounds using methods like LAMA or OpenCV-based techniques.

---

Installation

Prerequisites

• Python [3.9, 3.10]
• Dependencies listed in requirements.txt

Clone the Repository

git clone https://github.com/Mathixx/INF573-MakeBenardDisappear.git
cd INF573-MakeBenardDisappear

Install Dependencies

pip install -r requirements.txt

---

Usage

1. Prepare Input Media:

   • Place your images or videos in the input_media/ directory.

2. Run the Pipeline: It is mandatory to specify the type of media being processed using the --type option (video, photo, or live).

   Example for video processing:

   python main.py --type video --input input_media/example_video.mp4 --output output_media/

   Alternatively, you can modify the default_input and default_output_folder variables directly in main.py.

3. Additional Options:
   You can specify debugging levels and other options via command-line arguments:

   python main.py --type video --input input_media/example_video.mp4 --output output_media/ --debug_frames Detector --debug_video All

   Available Options:

    --input: Path to the input file (ignored for --type live).
    --output: Folder for processed output files.
    --debug_frames: Debugging level for individual frames. Options: None, Detector, Segmentor, Remover, All, complete_detector.
    --debug_video: Debugging level for video outputs. Options: None, Detector, Segmentor, All.
   

4. Visualize Results:

   • Output media will be saved in the output_media/ directory.

License

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

---

For more details, refer to the project repository.