This project aims to help the camera operators by making a system for video summarization. A video summarization algorithm detects essential events from the surveillance stream and can help index and efficiently retrieve required data from massive datasets. In this project, an efficient CNN is proposed that extracts key frames from long-length surveillance videos resulting in a summarized video. The model was then tested on the Raspberry Pi which is a resource-constrained device.