Portfolio of Md Touhidul Islam

Solution:

We developed SpaceXMag (Fig. 1), an optimization-based tool that automatically compacts whitespace in smartphone app interfaces while preserving layout structure. The system:

• Operates on DOM-based UI metadata, not pixels.

• Uses nonlinear optimization (GEKKO/IPOPT) to restructure UI layouts while maintaining element constraints.

• Compacts screens in a bottom-up hierarchical traversal, enabling local flexibility and global consistency (see Fig. 17).

• Supports three magnifier modes: fullscreen, window, and fisheye (see Figs. 1c, 1d, 1e).

• Includes a rendering backend that reconstructs modified layouts for visual output.

We conducted a user study with 11 low-vision participants and tested Space-Mag at scale on 16K+ screens from the RICO-SCA dataset.

Outcome:

• SpaceXMag is 28.13% faster in the overview task and 42.89% in the target acquisition task than baseline magnifiers.

• Users reported increased screen awareness, reduced zooming effort, and higher satisfaction, especially in window magnification mode.

• Scaled to compact over 16,500 real app screens, saving an average of 47.17% whitespace, with a median runtime of just 1.44 seconds per screen, making it up to 60x faster than prior solutions like SUPPLE.

Impact:

Space-Mag introduces a practical and scalable UI transformation technique for screen magnification tools. Its contributions include:

• Improving usability for low-vision users by significantly reducing navigation time and context loss.

• Demonstrating the power of DOM-aware, optimization-driven layout restructuring for real-world accessibility.

• Setting the stage for smarter, adaptive accessibility tools across platforms and user populations.