BoXnLabelS 0.1.0

A customizable image augmentation tool for object detection

BoXnLabelS

BoXnLabelS is a Python package for easy and customizable image augmentation, designed to generate augmented images and adjust their corresponding bounding box labels for deep learning models.

Installation

Install via pip:

pip install BoXnLabelS

Quick Start

import BoXnLabelS as bls

# Initialize the augmentation object
augmentor = bls.Image_Custom_Augmentation(
    SP_intensity=0.2,  # Salt & Pepper noise intensity
    CWRO_Key=20,       # Clockwise rotation in degrees
    CCWRO_Key=20,      # Counterclockwise rotation in degrees
    Br_intensity=True, # Brightness adjustment
    H_Key=True,        # Horizontal flip
    V_Key=True,        # Vertical flip
    HE_Key=True,       # Histogram equalization
    GaussianBlur_KSize=5,  # Gaussian blur kernel size
    Random_Translation=True,  # Random translation
    Scaling_Range=(0.75, 1.25),  # Scaling range (min, max)
    Img_res=540  # Image resolution
)

# Apply augmentations to a dataset
augmentor.Generate_Data(input_path="input_directory", output_path="output_directory")

Features

• Noise Addition: Salt & Pepper Noise
• Image Enhancements: Histogram Equalization, Brightness Adjustment
• Transformations: Rotation, Flipping, Translation, Scaling
• Blurring: Gaussian Blur
• Bounding Box Handling: Automatic YOLO format adjustment

More to come ...

Notes

• The module is under active development.
• Accepts images in JPG format only.
• Handles both labeled images (with bounding boxes) and unlabeled background images.