Real-time Object Detection and Tracking with YOLOv8 & Streamlit

This repository is an extensive open-source project showcasing the seamless integration of object detection and tracking using YOLOv8 (object detection algorithm), along with Streamlit (a popular Python web application framework for creating interactive web apps). The project offers a user-friendly and customizable interface designed to detect and track objects in real-time video streams from sources such as RTSP, UDP, and YouTube URLs, as well as static videos and images.

Explore Implementation Details on Medium (3 parts blog series)

For a deeper dive into the implementation, check out my three-part blog series on Medium, where I detail the step-by-step process of creating this web application.

WebApp Demo on Streamlit Server

Thank you team Streamlit for the community support for the cloud upload.

This app is up and running on Streamlit cloud server!!! You can check the demo of this web application on this link yolov8-streamlit-detection-tracking-webapp

Note: In the demo, Due to non-availability of GPUs, you may encounter slow video inferencing.

Tracking With Object Detection Demo

Tracking-With_object-Detection-MOV.mov

Overview

Yolov8.Streamlit.Rampal.Punia-1.mov

Demo Pics

Home page

Page after uploading an image and object detection

Segmentation task on image

Requirements

Python 3.6+ YOLOv8 Streamlit

pip install ultralytics streamlit pytube

Installation

Clone the repository: git clone https://github.com/CodingMantras/yolov8-streamlit-detection-tracking.git
Change to the repository directory: cd yolov8-streamlit-detection-tracking
Create weights, videos, and images directories inside the project.
Download the pre-trained YOLOv8 weights from (https://github.com/ultralytics/assets/releases/download/v0.0.0/yolov8n.pt) and save them to the weights directory in the same project.

Usage

Run the app with the following command: streamlit run app.py
The app should open in a new browser window.

ML Model Config

Select task (Detection, Segmentation)
Select model confidence
Use the slider to adjust the confidence threshold (25-100) for the model.

One the model config is done, select a source.

Detection on images

The default image with its objects-detected image is displayed on the main page.
Select a source. (radio button selection Image).
Upload an image by clicking on the "Browse files" button.
Click the "Detect Objects" button to run the object detection algorithm on the uploaded image with the selected confidence threshold.
The resulting image with objects detected will be displayed on the page. Click the "Download Image" button to download the image.("If save image to download" is selected)

Detection in Videos

Create a folder with name videos in the same directory
Dump your videos in this folder
In settings.py edit the following lines.

# video
VIDEO_DIR = ROOT / 'videos' # After creating the videos folder

# Suppose you have four videos inside videos folder
# Edit the name of video_1, 2, 3, 4 (with the names of your video files) 
VIDEO_1_PATH = VIDEO_DIR / 'video_1.mp4' 
VIDEO_2_PATH = VIDEO_DIR / 'video_2.mp4'
VIDEO_3_PATH = VIDEO_DIR / 'video_3.mp4'
VIDEO_4_PATH = VIDEO_DIR / 'video_4.mp4'

# Edit the same names here also.
VIDEOS_DICT = {
    'video_1': VIDEO_1_PATH,
    'video_2': VIDEO_2_PATH,
    'video_3': VIDEO_3_PATH,
    'video_4': VIDEO_4_PATH,
}

# Your videos will start appearing inside streamlit webapp 'Choose a video'.

Click on Detect Video Objects button and the selected task (detection/segmentation) will start on the selected video.

Detection on RTSP

Select the RTSP stream button
Enter the rtsp url inside the textbox and hit Detect Objects button

Detection on YouTube Video URL

Select the source as YouTube
Copy paste the url inside the text box.
The detection/segmentation task will start on the YouTube video url

movobjdetyoutubeurl.mov

Acknowledgements

This app uses YOLOv8 for object detection algorithm and Streamlit library for the user interface.

Disclaimer

This project is intended as a learning exercise and demonstration of integrating various technologies, including:

Streamlit
YoloV8
Object-Detection on Images And Live Video Streams
Python-OpenCV

Please note that this application is not designed or tested for production use. It serves as an educational resource and a showcase of technology integration rather than a production-ready web application.

Contributors and users are welcome to explore, learn from, and build upon this project for educational purposes.

rampal-punia/yolov8-streamlit-detection-tracking