Scribd
Upload
67 views3 pages

CV Report

This document describes a system for real-time human tracking and counting using YOLOv8, Supervision, and ByteTracker. YOLOv8 is used for fast and accurate object detection in videos. Supervision manages video data and coordinates detection and tracking. ByteTracker tracks detected humans across frames using unique IDs. The system can analyze prerecorded videos or live webcam feeds, detecting and counting humans while displaying output. It is a powerful and versatile computer vision system for security, planning, and business applications.

Uploaded by

humaragpt
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
67 views3 pages

CV Report

This document describes a system for real-time human tracking and counting using YOLOv8, Supervision, and ByteTracker. YOLOv8 is used for fast and accurate object detection in videos. Supervision manages video data and coordinates detection and tracking. ByteTracker tracks detected humans across frames using unique IDs. The system can analyze prerecorded videos or live webcam feeds, detecting and counting humans while displaying output. It is a powerful and versatile computer vision system for security, planning, and business applications.

Uploaded by

humaragpt
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Real Time Human Tracking and Counting Using

YOLOv8
Abhinav Raj(142202026) , Sanket D Avaralli (142202024)
December 2, 2023

Introduction
Our project introduces a different way to detect, track, and count people using computer vision. We
use three main tools: YOLOv8, which is great at spotting objects (like people) quickly and accurately;
Supervision, a system that helps manage and process video data; and ByteTracker, which is skilled
at keeping track of where objects move in a video. Together, these tools work on both videos that
have already been recorded and live video from webcams. This makes our system very flexible
and useful for many different situations.Click here to get the folder of the code file and the input-output
video.

1 YOLOv8 - Advanced Object Detection


1.1 Overview
YOLOv8 is a top-notch tool for finding and recognizing objects in real-time. It’s the newest version in
the YOLO series, known for being fast and accurate. It looks at an image by dividing it into many small
parts and figuring out what’s in each part at the same time. This way, it can find objects quickly and
efficiently.

1.2 Working Mechanism


• YOLOv8 processes the entire image in a single evaluation, making it exceptionally fast for real-time
applications.

• It uses deep convolutional neural networks to analyze the image, detecting objects and their loca-
tions.
• Advanced algorithms within YOLOv8 allow it to distinguish between various object classes, par-
ticularly focusing on human figures in this project.

1.3 Role of YOLOv8 in Our Project


In our project, YOLOv8 is the cornerstone for detecting humans in video frames. It provides critical
data like bounding box coordinates and class probabilities, forming the basis for tracking and counting
processes.

2 Supervision
2.1 Overview
Supervision is a comprehensive video processing framework designed to enhance the integration and effi-
ciency of object detection and tracking systems. It acts as a versatile toolkit, facilitating the management
of various video formats and sources. The framework excels in coordinating the flow of data between
object detection models, like YOLOv8, and tracking algorithms such as ByteTracker. It achieves this
by handling the annotation of video frames with detection and tracking data, making the output more
insightful and visually informative. Supervision’s key strength lies in its ability to seamlessly synchronize
the detection and tracking outputs, enabling a coherent and effective analysis of video content. This

1
synchronization is crucial for tasks that require detailed and accurate object tracking over time, such
as in surveillance and behavioral analysis. Supervision’s role is pivotal in ensuring that complex video
processing tasks are executed smoothly, maintaining the integrity and utility of the processed video data.

2.2 Role of Supervision in our Project


Supervision enhances the project’s functionality by managing video data efficiently. It serves as the
backbone for integrating YOLOv8’s detection outputs with ByteTracker’s tracking, ensuring a smooth
workflow in video processing.

3 ByteTracker
3.1 Overview
ByteTracker is an advanced tracking algorithm, highly effective in dynamic video environments. It
works seamlessly with object detection models, like YOLOv8, to track objects across video frames. The
algorithm assigns unique identifiers to each detected object, ensuring consistent and accurate tracking.
It utilizes a combination of feature extraction, Kalman filtering, and the Hungarian algorithm for data
association. This allows ByteTracker to predict the movement of objects, handle occlusions effectively,
and re-identify objects even after they have temporarily disappeared from the scene. Its strength lies
in robustness in crowded settings and real-time performance, though its effectiveness is contingent on
the quality of initial object detections. Overall, ByteTracker is a versatile and powerful tool for object
tracking in complex and dynamic scenarios.

3.2 Role of ByteTracker in our Project


ByteTracker plays a pivotal role in sustaining the tracking of humans across video frames. It uses the
detection data from YOLOv8 to assign unique tracking IDs, which is crucial for the accurate human
counting mechanism of the system.

4 Video-Based Tracking
4.1 Overview
In this model, we have taken a prerecorded video as our input and drawn a line by defining the coordinates
based on the requirements of the video.This line helps the model keep track of the number of humans
who have crossed the defined line in the video.

4.2 Process Explanation


1. Model : We took a pre-trained YOLOv8x.pt model for the detection of human in the video.
2. Frame Analysis: As the video plays, each frame (a single image from the video) is analyzed.
YOLOv8 identifies humans and draws bounding boxes around them.
3. Tracking with ByteTrack: The script then uses ByteTrack to keep track of each person. Byte-
Track assigns a unique ID to each individual, allowing the system to follow them across different
frames.
4. Counting Individuals: A tally is kept of the unique individuals seen throughout the video and
also ensures that the same person is not counted more than once.
5. Video Output:The output is an improved version of the original video, in which each individual
is distinctively marked and tracked using a unique ID. Additionally, the system separately counts
and displays the number of people crossing from each side of the line in the output video.

2
5 Live Webcam-Based Tracking
5.1 Overview
In this model, we are using live video from webcam as our input. Using YOLOv8 and ByteTracker, we
are detecting humans and tracking them by assigning a unique ID for each detected human and count
the number of different humans in the video, which is displayed in the live output video.

5.2 Process Explanation


1. Model Setup for Live Feed: Similar to the first model, YOLOv8 is used for detecting people.
However, it’s optimized for live webcam feeds.
2. Real-Time Frame Analysis: The model processes the live video feed, frame by frame then
YOLOv8 detects humans in real-time and creates bounding boxes around them.

3. Continuous Tracking: ByteTrack assigns unique IDs to each detected individual, enabling the
model to track people continuously as they move within the camera’s view.
4. Live Counting: The model counts the number of unique individuals detected over time. This
count is updated live, reflecting the current number of people seen by the webcam.

5. Live Video Output: The live feed is displayed with tracked individuals marked. A counter shows
the number of unique individuals detected in real-time.

Conclusion
This project successfully combines YOLOv8, Supervision, and ByteTracker to create a powerful com-
puter vision system. It can not only spot and follow people in real-time videos but also count them
accurately. This system is versatile and useful in many different areas, including security, city planning,
and business. Whether it’s working with prerecorded videos or live feeds from webcams, our project
shows how advanced technology can be used in practical, real-world situations.

You might also like