Video Data Augmentation

This code provides video data augmentation functionality using OpenCV and NumPy, designed to integrate into larger projects that require video preprocessing. It performs brightness adjustments and flip transformations to create an augmented dataset.

Code Overview

The main functionalities of this code include:

• Brightness Adjustments: Applies three levels of brightness augmentation to enhance data variability.
• Flip Transformations: Generates horizontally and vertically flipped versions of each video.

The code is organized with the following main components:

1. BrightnessAdjuster class: Uses the HSV color space to adjust the brightness of video frames at defined levels.
2. Data Augmentation Functions: Contains the brighten function, which performs pixel intensity clipping for brightness control.
3. Batch Processing in main_augment.py: Uses a list of directories and processes videos within each directory by applying augmentations (brightness and flips).

File Structure

• VideoProcessor.py: Contains classes and methods that implement augmentation logic, including brightness adjustment and flipping.
• BrightnessAdjuster.py: Defines the BrightnessAdjuster class for brightness augmentation.
• main_augment.py: Script for running batch processing on directories of videos, applying all augmentations specified in VideoProcessor.
• perform_augmentation.sh: SLURM job script for submitting jobs on HPC systems, handling environment setup and job submission.
• requirements.txt: Lists required Python packages (OpenCV and NumPy).
• README.md: Documentation for setup and usage (this file).

Setup and Installation

1. Install Dependencies

The following dependencies are required for this code:

• OpenCV (opencv-python-headless) for video processing.
• NumPy for numerical operations.

Install the dependencies from requirements.txt:

pip install -r requirements.txt

2. Configure Paths in main_augment.py

In main_augment.py, specify the directories containing the videos to be processed by adding them directly to the directory list varible.

For each directory listed, augmented videos will be saved in a corresponding output folder named _augmented.

Example configuration in main_augment.py:

directories = [ "/path/to/your/video_directory1", "/path/to/your/video_directory2", ]

3. Run the Code

Execute main_augment.py to apply brightness adjustments and flips to all videos within each specified directory.

python main_augment.py

Augmentation Techniques

Brightness Adjustment

The BrightnessAdjuster class and brighten function perform brightness adjustments. Within each video processing loop:

• Three Brightness Levels: The code applies three distinct brightness levels to each video, creating multiple brightness-augmented versions.

Flip Transformations

Both horizontal and vertical flips are applied using cv2.flip:

• Horizontal Flip: Creates a horizontally mirrored version of each video.
• Vertical Flip: Generates a vertically mirrored version.

Example Output

Augmented videos are saved in the output directory with naming conventions that specify each applied transformation:

• Brightness Levels: *_bright1.mp4, *_bright2.mp4, *_bright3.mp4
• Flip Transformations: *_flip_h.mp4 (horizontal) and *_flip_v.mp4 (vertical)

Optional: Running with SLURM

For HPC environments, install_packages.sh and perform_augmentation.sh automate environment setup and running the augmentation job.

Submit the job with:

sbatch install_packages.sh

sbatch perform_augmentation.sh