Beiwe

Your guide through the darkness. A chrome extension to block images in webpages based on user queries

Comment

Inspiration

Based on diagnostic interview data from National Comorbidity Survey Replication (NCS-R) on the prevalence of PTSD(Post Traumatic Stress Depression) among U.S. adults aged 18 or older:

> An estimated 3.6% of U.S. adults had PTSD in the past year
> Past year prevalence of PTSD among adults was higher for females (5.2%) 
        than for males (1.8%).
> The lifetime prevalence of PTSD was 6.8%.

Certain triggers can set off your PTSD, which switches your brain to danger mode. This may cause you to become frightened and your heart to start racing. The sights, sounds, and feelings of the trauma may come rushing back.

About 19 million Americans have one or more phobias that range from mild to severe. Phobias can happen in early childhood. But they are often first seen between ages 15 and 20. An estimated 9.1% of U.S. adults had specific phobia in the past year. Women are less likely to get treatment for phobias than men.

Several famous personalities like Johnny Depp, Daniel Radcliffe and William Shakespeare have suffered from phobias. Celebrities like Oprah and Tracy Morgan also suffer from PTSD. Even in our personal life, we have witnessed even close friends/relatives suffering due to it.

What it does

We propose Beiwe, a browser extension to browse without fear. Beiwe uses Image classification ML Models(ex YoloV3/Vision AI) to block images using user labels.

Let's say Aditya has coulrophobia (fear of clowns). When he sets "clown" as a label in Beiwe, every image containing a clown is blocked on every webpage he visits.

Further scaling can be achieved to utilise Video Object Detection to analyse and block/retrieve timestamps of undesired content in user-provided videos in the extension or in a separate WebApp. Face Detection using pre-trained face recognition models(FaceNet)  blocks images through user-uploaded photographs(ex: Block all images with Hitler). Feature detection and extraction using ORB/SIFT can help compare best-matched keypoints utilizing user-uploaded images(ex: Block all images having BMW logo).

Beiwe is built on a central database(MongoDb NoSQL DB) mapping images and labels to achieve blazing fast speed. So for example, if Neha also has coulrophobia, stored information in the database based on a previous visit by Aditya is used to block images which would result in a five-fold increase in performance! Cookie storage for previously visited webpages helps add another layer of performance enhancement. Local URL hashing in the browser itself ensures privacy.

How we built it

Tech Stack

HTML,CSS,JS for extension popup Flask ,Python for BackEnd Heroku Hosting Services MongoDB NoSQL Database Image Detection Model(YoloV3) Development Google Cloud Vision API NLTK NLP Library For Wordnet ORB/SIFT (Feature Extraction and Detection) Google Cloud Storage(in Development) Word2Vec Model(in Development) Docker(in Development) Fast.ai(in Development)

My teammate, Ganesh handled the front-end while I handled the backend and database development. It was challenging to decide on features to add and integrate our works at the end, but we were able to pull it off to a certain degree of success.

Challenges we ran into

  • Integrating our indivudal works
  • Deciding on features to add and finalizing design for the features
  • Researching the feasability of the feature and testing out prototypes of it
  • Integrating the basic functionality to our pre-built wireframe
  • Shifting Database to MongoDB from sqlite in production
  • Shifting from YOLOV3 pretrained model to Google CV
  • Testing out basic functionalities and fixing bugs and adding improvements
  • Communicating over long-distance with very little interaction
  • Splitting Tasks and keeping track of progress and issues to deal with through Notion
  • Learning Tech required to implement them on the fly with no previous experience handling them
  • Juggling personal responsibilites along with the Hackathon

Accomplishments that we're proud of

Technical

  • Minimal size and ram usage compared to other extensions in the market.

  • User-friendly and attractive UI greatly improves User Experience compared to other solutions.

  • Only in the recent past have mental health issues been taken seriously. Thus, very few products address the growing concerns over it (especially after COVID) that Beiwe tackles.

  • Beiwe is unique for its scalability which ensures enhanced performance with an increase in usage and users. (Central Database + Local Caching)

  • It stands out for its user-driven enhancement by allowing users to manually block and report images that can be verified and hidden in the future for new visitors to the webpage.

  • It also stands out for its privacy features, which ensure user browsing data isn't tracked and user labels are secured. Local URL Hashing + Local Storage of Labels

Non Technical

  • Overcoming the communication gap and splitting work efficiently
  • Learning and implementing a tech stack with very little previous experience
  • Coming up with a creative solution to a common problem with a decent working prototype
  • Gaining Experience using version control to maange our workflow

What we learnt

Technical

  • Gained experience in using Flask,MongoDB,Image Processing Models,Docker
  • Gained Experience using Figma
  • Learnt to Deploy a working model on Heroku
  • Tested and Experimented with Docker
  • Learnt Git/GitHub and gained experience in version control
  • Learnt Document Based DB and Relational DB and gained experience using them and integrating them with Python using ORM
  • Testing and integrating into a functioning prototype
  • Bug Fixing and Patching

Non Technical

  • Learnt to work as a team and manage time
  • Learnt to use Notion and other workflow management tools to efficiently manage and tackle issues
  • Learn to communicate and work together to achieve a common goal
  • Learnt Time Management
  • Developed practical skills
  •  Doing this project to make an impact in the real world is also something we are proud of since we grew up from making calculators and number-guessing games as projects : )

What's next for Beiwe

  • Scaling and Enabling easy deployment through Docker and GitLabs
  • Putting Beiwe on Chrome Extension Store
  • Porting the extension to other browsers (Safari,Firefox etc)
  • Adding new functionality like:
    • User Driven Enhancement
    • Customer Support with Person Well Being Chat Bot to interact with during troubling moments
    • Google Authentication
    • Batchwise Image Annotation
    • Handling Infinite Scroll in Social Media Websites
    • Utilise Video Object Detection to analyse and block/retrieve timestamps of undesired content in user-provided videos in the extension or in a separate WebApp.
    • Face Detection using pre-trained face recognition models(FaceNet)  blocks images through user-uploaded photographs(ex: Block all images with Hitler).
  • Creating a Community of Developers to work and further enhance the extension
  • Shifting the model to the extension side to further enhance privacy features

Built With

Share this project:

Updates