AccessLens

Inspiration

We built AccessLens after recognizing how dependent modern classrooms are on visual learning materials such as slides, whiteboards, and handwritten notes. While these formats are effective for many students, they create a significant barrier for visually impaired learners who cannot access information in real time. We wanted to reduce this dependency and enable a more independent learning experience. Our goal was to design a system that can convert visual lecture content into accessible audio, allowing students to follow along without needing someone else to interpret the material for them.

What it does

AccessLens is an accessibility-focused system that transforms classroom images into spoken explanations. A user simply captures an image of lecture content, and the system: Extracts text using Optical Character Recognition (OCR) Processes and simplifies the extracted text using AI Converts the final output into natural speech using text-to-speech In practice, this allows visual lecture material to be understood audibly, making classroom content more accessible and inclusive.

How we built it

We built AccessLens as a modular pipeline where each component handles a specific stage of processing.

The workflow is: The user captures an image using a camera input The image is passed into an OCR module that extracts raw text The extracted text is processed by an AI component that rewrites it into a simpler and more understandable format The processed text is converted into speech using a text-to-speech engine The system outputs both readable text and audio through a simple backend workflow

We structured the system so that each module operates independently, making it easier to debug, test, and extend.

Challenges we ran into

One of the main challenges was ensuring that all components worked reliably together as a single pipeline.

OCR results varied depending on image quality, lighting conditions, and text clarity, which sometimes affected downstream processing. We addressed this by improving preprocessing and handling cases where no text is detected.

Another challenge was integrating multiple technologies (OCR, AI processing, and speech synthesis) while maintaining a stable and predictable flow of data between them.

What we learned

Through building AccessLens, we learned how to design and integrate multiple AI systems into a single functional pipeline.

We gained experience in:

Optical Character Recognition (OCR) and image processing Structuring modular backend systems Connecting AI processing with real-world applications Handling edge cases such as poor image quality and missing text Designing technology with accessibility in mind

We also learned that building reliable systems is more important than adding complexity, especially when working under time constraints.

What's next for AccessLens

In the future, we aim to improve AccessLens by expanding its capabilities beyond static image input.

Possible future improvements include:

Better real-time processing for faster feedback Improved OCR accuracy under low-quality conditions Multilingual support for broader accessibility Smarter AI explanations tailored to different learning levels Cloud deployment for wider access across devices

Our long-term goal is to evolve AccessLens into a practical accessibility tool that can support real classroom environments and improve learning independence for visually impaired students.

Built With

• flask
• html
• javascript
• ocrengine
• openaiapi
• opencv
• python
• pyttsx3
• text-to-speech