Complete Jupyter Notebook can be accessed here : Public: Semantic Segmentation of YouTube Videos
Simply run the cells from top to bottom. Run the gradio app and enter any URL to a YouTube Video to obtain its semantically segmented transript!
The objective is to extract high-quality, meaningful (semantic) segments from any YouTube video.
In general, perofrm Semantic Segmentation of any YouTube video and use the transcript to align AI and ML models
Final Architecture used to perform the task
Here is a sample plot of the similarity of neighboring segments over the raw WhipserX transcript of a randomly chosen YouTube video