Waveshift Augmentation (Updated Version)

This is the updated version of the README, reflecting improvements to the Waveshift augmentation technique. The latest version, Waveshift WS 2.0, introduces an additional hyperparameter allowing control over the aperture, further enhancing its flexibility in data augmentation.

Table of Contents

• Built Modular Functions
• Installation
• Usage
• Approach
• Acknowledgements

Built Modular Functions

• CCWind: Square-crops an image to fit CNN architecture constraints (optional for other uses).
• FT2Dc: Applies a Fourier transform to the image.
• IFT2Dc: Applies an inverse Fourier transform to the image.
• PropagatorS: Constructs the wavefront at a given z-distance (WS 1.0).
• PropagatorPSF: Constructs the wavefront at a given z-distance with an adjustable aperture (WS 2.0).

Approach

This augmentation simulates the approximated propagation of light emitted by the target (e.g., a leaf) as spherical waves forming along the Z-direction (wavefronts). The WS 1.0 technique simulated camera shifts along these wavefronts. The updated WS 2.0 introduces an additional hyperparameter to control the aperture size, allowing for more refined adjustments to light properties captured in the augmented image.

Installation

To install the required dependencies, run:

pip install -r requirements.txt

Usage

Here's how to integrate our data augmentation technique into your PyTorch data pipeline:

import torch
from torchvision import transforms
import numpy as np
from PIL import Image, ImageFile
from transforms import CCWind, FT2Dc, IFT2Dc, PropagatorS, PropagatorPSF
from Waveshift import Wavefront_Shift
import matplotlib.pyplot as plt
import numpy as np

# Define the transformation pipeline
waveshift_transform = Wavefront_Shift(mode_="s")
transform_pipeline = transforms.Compose([
    transforms.Resize((512, 512)),
    waveshift_transform
])

Load an image and apply the Waveshift augmentation. The propagator construct is determined by a random number from 1 to 41m, and the aperture radius is set to 0.01 by default.

# Load an image
img_path = 'test.JPG'  # Replace with your image path
img = Image.open(img_path).convert('RGB')

# Apply transformations
transformed_img = transform_pipeline(img)

Visualize the original and augmented images side by side:

# Display the original and transformed images
plt.figure(figsize=(10,5))

plt.subplot(1,2,1)
plt.imshow(img)
plt.title("Original Image")
plt.axis('off')

plt.subplot(1,2,2)
plt.imshow(transformed_img)
plt.title("Augmented Image")
plt.axis('off')
plt.show()

For a clearer comparison, plot the difference image:

# Resize images to the same size
new_size = (512, 512)  # use transforms.Resize for matching
img_resized = img.resize(new_size)

# Convert resized images to numpy arrays
img_np = np.array(img_resized)
transformed_img_np = np.array(transformed_img)

# Calculate the absolute difference between the transformed image and the original image
diff = np.abs(transformed_img_np - img_np)

# Normalize the difference image to range [0, 1] for display (if needed)
diff_norm = diff / np.max(diff)

# Plotting the difference image
plt.figure(figsize=(10, 5))
plt.imshow(diff_norm, cmap='gray')  # Use 'gray' for grayscale images, or leave it for RGB
plt.title("Difference Between Original and Transformed Images")
plt.axis('off')
plt.show()

Exploring Extended Hyperparameter Configurations in WS 2.0

Beyond the physically meaningful regime of real, non-negative (z, R) values, we investigated a wider range of hyperparameter combinations to assess whether the WS 2.0 propagator maintains structural consistency under theoretical extensions. Figure A1 visualizes a systematic sweep through both real (positive and negative) and imaginary values.

• Sign-inverted configurations—such as (−z, R) and (z, −R)—retain both intensity structure and propagation phase in Fourier space, generating mirrored patterns that remain consistent with the original physical formulation.

• Imaginary-valued configurations—where z → jz and R → jR—yield distinct surface behaviors. Imaginary z values result in inverse propagators with characteristics akin to high-pass filtering. Conversely, imaginary R disrupts the diffraction structure, leading to unstable or incoherent patterns compared to the physically derived formulation.

While these configurations are not used in our practical augmentation pipeline, they illustrate the sensitivity and constraints of the WS 2.0 propagator. This underscores that its expressive power originates from its physical foundation, distinguishing it from arbitrary mathematical alterations that lack interpretability, stability, or practical benefit.

Original (z, R) Configuration	Sign-Inverted (z, −R) Configuration
Imaginary z (jz) and Real R	Imaginary z (jz) and R (jR)

Acknowledgements

This project is currently not licensed. The authors have published WS 1.0 in IEEE Access, and WS 2.0 is submitted to Electronics. For inquiries, please contact Gent Imeraj at [email protected].

• G. Imeraj and H. Iyatomi, "Waveshift Augmentation: A Physics-Driven Strategy in Fine-Grained Plant Disease Classification," in IEEE Access, vol. 13, pp. 31303-31317, 2025, doi: 10.1109/ACCESS.2025.3541780.

Keywords: Plant diseases, Adaptation models, Data models, Computational modeling, Cameras, Data augmentation, Wavefront shifting, Automated plant disease diagnosis.