TactileNet: Bridging the Accessibility Gap with AI-Generated Tactile Graphics for Individuals with Vision Impairment
Project Overview
The document discusses TactileNet, an innovative AI-driven framework aimed at generating tactile graphics for individuals with vision impairments, thereby enhancing educational accessibility. Leveraging a comprehensive dataset and advanced AI methodologies such as Low-Rank Adaptation (LoRA) and DreamBooth, TactileNet automates the creation of high-quality tactile images, significantly improving the educational experience for visually impaired learners. The framework not only streamlines the design process of tactile graphics but also demonstrates a strong commitment to addressing the accessibility gap in education. Overall, TactileNet showcases the potential of generative AI to transform educational resources, making them more inclusive and effective for all learners, particularly those with disabilities.
Key Applications
TactileNet - a dataset and AI-driven framework for generating tactile graphics
Context: Accessibility in education for visually impaired learners
Implementation: Utilizes text-to-image Stable Diffusion models fine-tuned with LoRA and DreamBooth to create tactile graphics from text prompts.
Outcomes: Achieved 92.86% adherence to accessibility standards, demonstrating high fidelity and structural similarity to expert-designed tactile images.
Challenges: Challenges include the scarcity of high-quality paired datasets, the complexity of tactile graphic requirements, and the need for further refinement of generated images.
Implementation Barriers
Technical Barrier
Scarcity of paired datasets for training effective AI models for tactile graphic generation.
Proposed Solutions: Developing a comprehensive dataset (TactileNet) that integrates high-quality tactile images and textual descriptions.
Implementation Barrier
High costs associated with refreshable tactile displays and the complexity of tactile graphic design, leading to challenges in production.
Proposed Solutions: Automating tactile graphic generation to reduce dependency on manual design, thus lowering production costs.
Project Team
Adnan Khan
Researcher
Alireza Choubineh
Researcher
Mai A. Shaaban
Researcher
Abbas Akkasi
Researcher
Majid Komeili
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Adnan Khan, Alireza Choubineh, Mai A. Shaaban, Abbas Akkasi, Majid Komeili
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai