¯\(ツ)/¯ Blocker

• 

  What you will see when you are not being productive!

Inspiration

Have you ever been trying to complete course work, but have ended up falling into temptation and clicking into a new tab and opening YouTube, TikTok or another website that negatively affects your productivity by taking advantage of our dopamine seeking brains? We sure have!

What it does

Our chrome extension leverages Computer Vision to automatically block non-productive websites while enabled, easing the burden on you, the user, by not having to mechanically filter through websites that may distract you!

How we built it

• Built TensorFlow model from scratch and fine-tuned a pre-trained model by freezing a base model and adding layers
• Collected data from scratch by automating a selenium script and classifying them in folders for training/testing
• Developed a React-based Chrome extension which sends requests to a Python FastAPI backend for classification, and blocks unproductive websites
• Utilized SQLite for caching and crowdsourcing classification, reducing duplicate work for the same screens

Challenges we ran into

• Smaller than required data pool using from-scratch convolutional layers, so we opted to use a pre-trained model for better results

Accomplishments that we're proud of

• 98% accuracy on training data, 80% accuracy on testing data
• Retains screen privacy, as it only screenshots what renders after redirects to unauthenticated screens

What's next for ¯\(ツ)/¯ Blocker

• Productionized deployment for large-scale use, such as per-user profile blocking
• Usage on mobile devices
• Rewards for users who stay productive

Built With

• css
• fastapi
• html
• javascript
• python
• react
• selenium
• sqlite
• tensorflow
• typescript