In this assignment, you are tasked with creating embeddings from video data and using these embeddings to classify videos into predefined categories. This project is structured to mimic the workflow required for a content recommendation system on a short video platform, where accurate categorization is crucial for content discovery and recommendation quality.
Writing code isn’t a requirement. We only expect you to present an approach for the task. Slides would save time during the interview and would show some commitment leading to some brownie points.
- Extract features from video files to create embeddings.
- Use the embeddings to classify videos into categories using machine learning techniques.
- Evaluate the performance of the classification models.
- Design both supervised and unsupervised approaches for the task.
- You are allowed to use open source models and data.
- Pipeline should be scalable for millions of users and videos.