Video Analytics System(VAS)

Introduction

This task aims to create a simple web-based Video Analytics System (VAS) to analyze content from videos with the following inputs and outputs:

Input:
- Any YouTube video URL
- Upload a video from a local computer
Ouput: Analyzing basic video content and display results on the web while playing the video. For example:
- Counting people and other objects in each frame and presenting the results synchronized while playing the video;
- Illustrating a canvas that exhibits detected objects (bounding box and object name) in each frame, synchronized with the video.

1. Approach

In this task, I utilize Flask, a Python framework, to develop a basic website, while employing YOLOv8 model for content analysis in videos. Furthermore, in order to handle Youtube URLs, I employ pytube library for downloading YouTube Videos. Notably, all results in the Example results section were obtained using the nano YOLOv8 model(yolov8n.pt). Furthermore, if you have a strong GPU, you may consider using a larger YOLOv8 model, which provides more accurate predictions but takes longer to process.

2. To run VAS

I recommend creating an anaconda environment:

conda create --name vas python=3.9

Then, install Python requirements:

pip install -r requirements.txt

Finally, from the vas project root, run:

python app.py

3. Overview

i. UI

This section illustrates the website's layout and its functions

Home page:
Upload page:

ii. Example results

This section provides an overview of the results from the provided examples located in the Examples folder. Furthermore, please access the complete video outputs by following this link.