Pixeli

Automatic video summarization and indexing with deep learning

Comment

Inspiration

We were inspired by the rise of video content, especially in education, and how there was no easy way to find information quickly or get the TLDR without having to watch the video through.

What it does

Pixeli is an easy to use website that allows users to choose from any Youtube video, then provides a way to find all parts of the video where a certain word/sentence is said and easily jump to it, also providing a short summary of the video content.

How we built it

We used Tensorflow along with an OCR and audio transcription to. We used natural language processing to create a short summary of the video content. We used Rust for the backend of the website and Next.js/React to create the frontend.

Challenges we ran into

We ran into some challenges with building the machine learning model was that it wasn't fast enough. Specifically, the OCR was too slow so we used our own custom fuzzy search algorithm.

Accomplishments that we're proud of

We're proud of the ease of use and power behind the project we built as well as the things we've learned during our time in Hack the North.

What we learned

We learned alot about machine learning, and frontend and backend web development.

What's next for Pixeli

We plan to expand Pixeli to allow for additional functionality to analyze and index videos and incorporate other forms of media too.

Share this project:

Updates