Youtube Thumbnail Analysis

Inspiration

To solve the problem of determining what makes a successful thumbnail in a given category, our multi-label classification analysis tool provides marketing agencies with the ability to quickly and accurately analyze the components of a successful thumbnail. By leveraging machine learning algorithms and data analysis techniques, our tool can identify key elements such as colors, text, and imagery that have been proven to be effective in a given category.

With our tool, marketing agencies can save time and resources by avoiding trial-and-error methods for creating thumbnails. Instead, they can use data-driven insights to create high-quality thumbnails that resonate with their target audience and increase engagement rates. This not only leads to improved marketing outcomes for the agency's clients, but it also helps to create a more engaging and user-friendly digital environment for viewers.

Our tool is also designed to be user-friendly and accessible, allowing even those without technical backgrounds to utilize its powerful capabilities. Additionally, as our tool continues to learn and adapt based on user data and feedback, it will only become more effective over time, further improving marketing outcomes and creating a better digital experience for all.

What it does

Request data using youtube API and gets 10 random gaming channels and gets their interaction count, publication date, and channel name. Use NLP to analyze successful video titles vs unsuccessful ones. We then attempted to perform computer vision on the thumbnails to detect features in successful youtube thumbnails

How we built it

Our code uses the YouTube API to request data for 10 random gaming channels, and excluded youtube shorts. We then saved each corresponding thumbnails into our google drive. Manually labeled thumbnails(if each thumbnails had a person, bright colors, or/and text). Grabbed the titles and separated the data into successful/unsuccessful videos