AccessPath

Main Screen
Allows you to choose which accessibilities you want to cater to!
Our own custom Deep Research Component, more in the video!
Depth image of the staircase outside our hackathon room!
Annotated image of the same staircase!

Inspiration

AccessPath was born from a profound recognition of the challenges faced by individuals with diverse accessibility needs in navigating public spaces, whether they have mobility limitations, visual or auditory impairments, cognitive differences, or other accessibility requirements. While many building owners and facility managers genuinely want to improve accessibility, they often find themselves overwhelmed by the technical complexities of implementing these changes. The gap between understanding what needs to be done and visualizing how these modifications would actually look in their space creates a significant barrier to action. We realized that by harnessing the power of advanced AI technology and combining it with established accessibility standards, we could bridge this crucial gap. Our inspiration came from the belief that technology should not just identify problems but should provide clear, actionable, and visually compelling solutions that make accessibility upgrades more approachable and implementable. By creating a tool that can transform a simple photograph into a comprehensive accessibility upgrade plan with realistic visualizations, we aim to accelerate the adoption of accessibility improvements and contribute to creating a more inclusive built environment that truly serves everyone, regardless of their abilities.

What it does

AccessPath is a revolutionary AI-powered system that transforms how we approach accessibility upgrades in public spaces. The platform begins by analyzing photographs of spaces that need accessibility improvements, leveraging computer vision to identify potential upgrade opportunities and generating comprehensive recommendations that strictly adhere to AODA standards and other regional standards. At its core, AccessPath creates realistic visualizations of these proposed upgrades using Gemini's cutting-edge multimodal capabilities, allowing stakeholders to see exactly how modifications would look in their space, complemented by detailed cost estimates and material recommendations derived from extensive market research. What sets AccessPath apart is its holistic approach to accessibility planning - rather than simply suggesting changes, it provides a complete roadmap for implementation, including material specifications, contractor recommendations, and maintenance requirements, ensuring that accessibility upgrades are not just theoretically sound, but practically implementable with multiple variations that balance aesthetics, functionality, and budget considerations.

How we built it

We developed a sophisticated end-to-end pipeline that processes images through multiple AI models to generate detailed accessibility upgrade recommendations and visualizations:

1. Image Preprocessing

Takes user-uploaded images
Enhances image quality through sharpening and contrast adjustment
Prepares the image for feature detection and depth analysis

2. Feature Detection & Spatial Analysis

Implements a spatial model to generate depth maps of detected features
Combines object detection with depth information to understand exact dimensions and spatial relationships

3. Accessibility Analysis (Gemini)

Feeds the preprocessed image and spatial data to Gemini's multimodal model
Generates detailed textual descriptions of recommended accessibility upgrades
Considers spatial relationships and dimensions from depth analysis

4. Structured Data Processing (Cohere)

Processes Gemini's output through Cohere's Command model
Enhances upgrade descriptions with specific details
Ranks different upgrade options based on priority and impact
Converts unstructured recommendations into actionable JSON format

5. Deep Research Component [No use of deep research api, rlly proud of that :)

Automatically researches costs for recommended upgrades
Gathers detailed information about materials and textures
Analyzes compliance requirements and installation specifications
Compiles comprehensive data about contractors and maintenance

6. Final Visualization (Gemini 2.0 Flash)

Takes all processed data (upgrades, materials, dimensions, costs)
Generates hyperrealistic visualizations of the space with proposed upgrades
Creates multiple variations based on different material and design options

This pipeline ensures that each recommendation is not only visually compelling but also practical, compliant with accessibility standards, and backed by detailed research and cost analysis.

Challenges we ran into

Ensuring accurate spatial understanding: Combining object detection with depth estimation to properly place accessibility upgrades
Material representation: Getting the AI to generate realistic textures and materials for different upgrade options
Maintaining consistency: Ensuring that generated visualizations accurately reflect the specific materials and dimensions from our research data
Processing efficiency: Implementing parallel processing to handle multiple upgrade variations without overwhelming API limits
Integration complexity: Coordinating multiple AI models (Gemini, Cohere) to work together seamlessly such that the efficiency holds up

Accomplishments that we're proud of

Created a fully automated pipeline from image input to hyperrealistic upgrade visualization
Successfully integrated Canadian accessibility standards (AODA) and other regional standards into the AI's decision-making process
Developed a sophisticated deep research component that provides real-world cost estimates and material options
Achieved high-quality visualizations of accessibility upgrades using Gemini 2.0 Flash
Built a system that can generate multiple variations of upgrades while maintaining visual consistency

What we learned

Throughout the development of AccessPath, our team gained invaluable insights into the intricacies of working with advanced AI systems and accessibility standards. We discovered the critical importance of structured prompting when working with multimodal AI models, learning how precise and well-crafted prompts significantly improve the quality and consistency of AI-generated outputs. This knowledge proved essential as we developed techniques for seamlessly integrating multiple AI services into a cohesive pipeline, requiring careful orchestration of different models and APIs to work together effectively.

Our work with spatial data presented unique challenges that pushed us to develop sophisticated methods for processing and analyzing complex architectural information. We learned to balance accuracy with performance, particularly when implementing parallel processing strategies to handle multiple API requests within rate limits. This technical expertise was complemented by a deeper understanding of accessibility standards and their practical implementation - a complex landscape of regulations, guidelines, and real-world considerations that must be carefully navigated to create truly effective accessibility solutions. These lessons have not only enhanced our technical capabilities but also deepened our appreciation for the intricate relationship between technology and accessibility design.

What's next for AccessPath

The future of AccessPath is focused on expanding its capabilities and reach to make accessibility planning more accessible and effective than ever before. We plan to enhance the system's spatial understanding by integrating real-time 3D modeling capabilities and AR visualization features, allowing users to see proposed upgrades in real-time through their device's camera. This will be particularly valuable for complex architectural spaces where precise measurements and spatial relationships are crucial, enabling even more accurate and detailed upgrade recommendations.

Looking ahead, we're developing a comprehensive database of successful accessibility upgrades, which will continuously improve our recommendation engine through machine learning. These enhancements, combined with our commitment to user-centered design and real-world implementation experience, will help AccessPath become the go-to platform for accessibility planning and visualization, ultimately contributing to a more accessible world for everyone. Through strategic partnerships with accessibility consultants and contractors, we'll continue to validate and refine our recommendations, ensuring that AccessPath remains at the forefront of accessibility innovation.