Inspiration
Equality is always established in all walks of human life. In the past two decades, social media has become an integral part of our lives. Today, an average young user spends about five hours of the day on social media. Social media drives decisions, emotions, and value. A medium that is so vastly used sees barely any inclusivity efforts for visually impaired users. It’s sometimes considered as unexplorable waters for these sections. In our efforts to eradicate such barriers ‘Dark Room’ is created.
Social media platforms allow users to post pictures along with alternate text options to have equal reach throughout the audience including the visually impaired. However, this feature has negligible penetration due to a lack of awareness among the users. So a good low-cost solution such as ‘Dark Room’ is potentially a viable solution that does not rely on zero adoption and is one hundred percent compliant.
A few movies in streaming media made in recent times include the video description with an intention to reach out to the visually impaired audience. However, the production houses are not fully working towards including such narrations for the older content. So with this product, the limitations of this issue can be mitigated and content can be made better accessible to the groups in discussion. Additionally, video and streaming platforms such as YouTube, Netflix, Hulu, and HBO Max can just plug in the application and make available audio descriptions to almost all content in a negligible time and minuscule investment.
What it does
Our platform helps the visually impaired access social media to stay connected, build relations, access information on world events, and find employment opportunities. Over time there has been a shift beyond comprehension in the interplay of human relations. This product promoted the visually impaired to keep in pace with the changes and enjoy the same and equal opportunities to grow and bond.
How we built it
Front end: We first developed our front end on React application that communicates with Python Flask API to successfully provide audio translation for each picture frame in the video.
Back end: In the next phase we used Redis to create our user and video data tables in the database.
Machine Learning: For image captioning Blip architecture was used to fetch descriptions for photos and video frames. The model was fine-tuned through active learning by training through many real-time data feeds.
Challenges we ran into
The first challenge that we faced while building our product was connecting the front end and back end. And another roadblock we had to deal with was interacting with the Redis DB. The other problem that we spent time on was reducing the processing time for the descriptions on the video.
Accomplishments that we're proud of
We are proud that through our product we can bridge the large unattended accessibility gap among people. We believe our product will create a large-scale impact in the lives of visually impaired people. Through this app, a visually impaired person who wishes to seek employment or start a small shop can have an easier way like all of us do.
What we learned
→ Redis
What's next for DARK ROOM
Going forward we want to integrate selective descriptions to ensure concise and confined narrations that align with the video’s original audio to improve viewer experience.
Log in or sign up for Devpost to join the conversation.