Pixel Puzzle

Inspiration

The inspiration behind Pixel Puzzle was the intriguing combination of technology and mystery. The goal of this challenge was to extract valuable information from a low-quality video. To accomplish this project, 3 steps were identified: 1) transcoding 2) restoration 3) number identification using AI

How we built it

Pixel Puzzle was made using a variety of programming tools. Our primary programming language was Python

We started by utilizing MP4Box for the transcoding process, converting the .h264 video format into a more versatile .mp4 format. For the task of video restoration, we leveraged OpenCV, a powerful computer vision library, to enhance and clarify the video frames. The number identification phase was driven by the Google Vision API, known for its strong machine learning capabilities in image recognition.

Additionally, Google Cloud served as our backbone for processing and storage. Our combination of tools allowed us to build an efficient solution for solving the challenge at hand.

Challenges we ran into

This project was out of all of our comfort zones. We learned more about video processing, as well as the nuances of machine learning in data extraction.

What's next for Pixel Puzzle

In the future, we want to improve Pixel Puzzle's AI capabilities for more extensive uses, such identifying various kinds of data patterns in videos. Additionally, we desire to optimize the video restoration procedure to be used with video inputs of even lower quality.

Built With

• google-cloud
• google-vision-api
• mp4box
• opencv
• python