Muhammad Zaigham Zaheer

Title: Muhammad Zaigham Zaheer and Label-Preserving Data Augmentation

Introduction

Muhammad Zaigham Zaheer is an inventor based in Abu Dhabi, United Arab Emirates. His work includes a patented approach to creating enriched training images with diffusion models, hybrid image construction, and fractal blending. He is associated with Mohamed Bin Zayed University of Artificial Intelligence.

Latest Patents

Zaheer is listed as an inventor on 1 patent, titled “System and Method for Label-Preserving Data Augmentation with Diffusion Models.” The invention describes a method for generating augmented training images. It uses an original image and one or more prompts to produce a corresponding image with a diffusion model.

The method then combines part of the original image with part of the generated image using a binary pixel-wise mask. The mask contains zero- or one-valued pixels and has at least one entire row or column set to zero. The original and generated images are multiplied by the mask and its complement, respectively, to form a hybrid image. Finally, a fractal image is blended into the hybrid image to create the augmented result.

Career Highlights

Zaheer’s listed patent focuses on data augmentation for machine-learning training. By combining original and generated image content with fractal blending, the disclosed system aims to create varied training images while preserving their labels.

Collaborations

Khawar Islam and Arif Mahmood are identified as Zaheer’s coworkers. His association with Mohamed Bin Zayed University of Artificial Intelligence places his work within an academic artificial-intelligence setting.

Conclusion

Muhammad Zaigham Zaheer’s patent presents a structured approach to label-preserving image augmentation that combines diffusion-generated content, masked image concatenation, and fractal blending.