Scribd
Upload
11 views3 pages

Review

YOLO Object detection

Uploaded by

j5nh2hpftp
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
11 views3 pages

Review

YOLO Object detection

Uploaded by

j5nh2hpftp
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 3

Slide 1

Dharssinee:
Our project title is object detection using yolov8

Im dharssinee

karthika:
And im karthika.

Reshma:
Im reshma

Slide 2(abstract)
Dharssinee:
Object detection is crucial in various domains including cctv cameras, self driving cars etc

The main objective of this experiment is to make the computers to identify and detect objects in the
visual data.

. This idea is used to identify multiple objects in images, videos and live webcam

We use yolo algorithm version 8 which is a special algorithm for object detection. And it is the latest
version released in 2023 by ultralytics.

Slide 3:(introduction)
Reshma:
YOLO has many versions such as YOLOV1 to yolov8.

YOLOv8 specifically is chosen for its capabilities in accurately detecting objects within visual data

We are all using mobile phones and it has the feature to unlock mobile using face recognition. The
unlock feature follows the computer vision technique as face recognition is an application object
detection. Object detection can be done by various algorithms and methods, here we have used
YOLOV8 (You Only Look Once Version 8) an effective real time object recognition algorithm

.Slide 4(highlights):
Karthika:
Our project Operates with three different input types: images, videos, and live feeds.

It has The ability to use both pre-trained models and custom models and capability to identify 85
different classes.
Slide 5(architecture diagram)
Karthika:
This is the architecture diagram

The process begins with datacollection

We collected most of the images from realtime. And downloaded some from google images.

And all images are imported in roboflow. And we used the help of the tools available to annotate
them

the dataset of images and labels are given as input for training

And then they are processed in the backbone cnn.

The backbone cnn extract the abstract information of the image like edges , textures , shapes and
strcutures. It generates feature maps

The feature maps are then processed through the feature pyramid network.

FPN helps YOLOv8 detect objects at different sizes in an image by creating a "pyramid" . and then it
adjusts the resolution of the feature maps.this helps yolov8 suitable for various applications

Reshma
BOUNDING BOXES

Then multiple yolo head process the features and predicts the bounding box for objects by
continuously predicting the parameters in each grid.

It calculates the confidence score that indicates the probability that the box contains the object

It also predicts the class probabilities

Dharssinee
Loss function:

The predictions are compared to the original annotations using a custom loss function

The error is given through the network and the model parameters are updated to minimize loss
function

After training …..the model is created

And that model is used for object detection

Slide 6:
Dharssinee:
This is how our program works

At first it will load the yolo model.

That is yolo .pt file which contains the weights of the trained dataset.
it can predict 85 classes

Karthika:
we have given our new customized dataset which contains the classes of eye, nose, mouth and our
college id card.

Reshma:
So after loading the model.

We are allowed to choose the type of input whether its an image, video or live

If it’s a image

It will read inout, process using yolo, predict and draw bounding boxes

If it’s a video

It will consider frames for each millisecond and detect objects for each frame and display it

If its live

It open the webcam

And read each frame , process , predict and display the output

Slide 8:
Reshma:
This is our sample code . and this is how our projects works and detects objects

You might also like