Laser Beam Synthetic Dataset: Gaussian, Super-Gaussian, Bessel, Multimode, and Speckled Beams
Description
This dataset, titled "Laser Beam Synthetic Dataset: Gaussian, Super-Gaussian, Bessel, Multimode, and Speckled Beams," contains 10,000 computer-generated laser beam images with various intensity distributions. These beams are widely used in optics, photonics, and laser physics research and serve as a valuable resource for machine learning-based beam classification, optical simulations, and laser system analysis. Dataset Composition: - 9,500 synthetic laser beam images generated using numerical simulations - 500 reference Gaussian beam images to serve as a standard baseline - Metadata file (`metadata.csv`) containing beam characteristics, including type and quality factor Each image is 256×256 pixels, grayscale (8-bit, PNG format), ensuring compatibility with computer vision models, deep learning frameworks, and scientific image analysis tools. Beam Types Included: 1. Reference Gaussian: Standard Gaussian beams used as a baseline 2. Gaussian Beam: Intensity follows a normal distribution, simulating laser beams with a single mode 3. Super-Gaussian Beam: Higher-order Gaussian beam with sharper edges, often used in high-power laser applications 4. Bessel Beam: Non-diffracting beam generated using Bessel functions, useful in optical trapping and microscopy 5. Multimode Beam: Superposition of multiple Gaussian beams to simulate real-world laser outputs with multiple transverse modes 6. Speckled Beam: Randomized intensity distribution mimicking speckle patterns found in scattered laser light Beam Quality Metrics: Each beam image is analyzed using multiple beam quality factors, stored in the metadata file: - Beam_Quality_Factor (M²): Measures beam propagation characteristics - Symmetry Score: Assesses left-right symmetry of the beam profile - Circularity Index: Evaluates the beam's shape compared to an ideal circle - Full-Width at Half Maximum (FWHM): Measures beam width for intensity profile analysis Applications of this Dataset: - Deep Learning & AI: Train neural networks for beam classification and segmentation - Optical System Design: Simulate and optimize laser systems based on different beam profiles - Laser Diagnostics: Develop automated tools for beam quality assessment - Biomedical Imaging: Enhance laser applications in medical imaging and microscopy - Computational Optics: Explore structured light, wavefront shaping, and non-diffracting beam propagation Why This Dataset? This dataset is designed to bridge the gap between theoretical optics and AI-based analysis. Unlike real-world laser beam datasets, which require expensive laboratory setups, this synthetic dataset provides a cost-effective, noise-controlled, and scalable solution for training and evaluating optical AI models. Researchers, engineers, and machine learning practitioners can use this dataset to test deep learning algorithms, validate optical theories, and develop automated beam analysis tools.
Files
Steps to reproduce
1. Data Generation Protocol To generate the dataset, five different laser beam profiles were simulated: - Gaussian Beam: Representing a standard single-mode laser beam with a radially symmetric intensity distribution. - Super-Gaussian Beam: A modified Gaussian beam with sharper intensity roll-off, used in high-power and precise laser systems. - Bessel Beam: A non-diffracting beam profile generated using the Bessel function, useful in optical trapping and microscopy. - Multimode Beam: A combination of multiple Gaussian beams, simulating a laser with several transverse modes, commonly found in real-world sources. - Speckled Beam: Generated by adding random noise to a Gaussian beam to simulate the intensity variations seen in speckle patterns. Each of these beam profiles was mathematically modeled and generated through numerical simulations. The models utilized the fundamental principles of beam propagation and diffraction, with noise added to ensure the beams closely resembled real-world laser outputs. 2. Image Generation and Processing The images in the dataset are 256×256 pixels, chosen for their balance between computational efficiency and visual detail. For each beam type, images were generated by simulating the laser’s intensity distribution and applying Gaussian noise to simulate imperfections and environmental noise. Additionally, each beam's quality metrics were computed: - M² (Beam Quality Factor): Measures how closely a beam approximates an ideal Gaussian beam, with higher values indicating more divergence. - Symmetry: The beam's symmetry was quantified by comparing the left and right halves of the beam profile. - Circularity: This metric measures the roundness of the beam, where a value of 1 indicates a perfect circle. - FWHM (Full Width at Half Maximum): The beam’s width was calculated based on its intensity profile. 3. Metadata Collection Each image in the dataset is associated with a metadata entry, stored in a CSV file. The metadata includes: - Filename: The unique identifier for the image file. - Beam Type: The beam type (e.g., Gaussian, Super-Gaussian, Bessel). - M² (Beam Quality Factor): A numerical measure of beam propagation. - Symmetry Score: A measure of the beam's symmetry. - Circularity Index: A measure of the roundness of the beam's intensity distribution. 4. Software and Tools Used The dataset was generated using the Python programming language, along with several key libraries: - NumPy: Used for numerical operations, array manipulations, and beam profile generation. - SciPy: Provided advanced mathematical functions, including Bessel functions for non-diffracting beams and Gaussian fitting routines. - OpenCV: Used for image saving and analysis, such as calculating moments to assess the circularity of the beams. - Pandas: Managed and saved the metadata associated with each beam in CSV format.