Scribd
Upload
15 views4 pages

Object Detection RTRP Report

The RTRP Project Report details the implementation of a real-time object detection system using the YOLOv5 architecture, trained on the COCO dataset. The project aims to accurately detect multiple object classes in various environments, achieving a mean Average Precision of 85.2% and an inference speed of 25 FPS. Future enhancements include mobile deployment and integration with robotics for specific applications.

Uploaded by

Panduranga Dhondi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
15 views4 pages

Object Detection RTRP Report

The RTRP Project Report details the implementation of a real-time object detection system using the YOLOv5 architecture, trained on the COCO dataset. The project aims to accurately detect multiple object classes in various environments, achieving a mean Average Precision of 85.2% and an inference speed of 25 FPS. Future enhancements include mobile deployment and integration with robotics for specific applications.

Uploaded by

Panduranga Dhondi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 4

RTRP Project Report: Object Detection Using Deep Learning

Project Title

Real-Time Object Detection Using YOLOv5

Team Members

- Name 1 (Reg.
No.)

- Name 2 (Reg.
No.)

- Name 3 (Reg.
No.)

- Name 4 (Reg.
No.)
Institution

[Your College/University
Name]

Department

[Department Name, e.g., Computer Science &


Engineering]

Duration

[Start Date] to [End


Date]

Abstract

Object detection is a fundamental task in computer vision, aiming to identify and locate objects within an

image or video. This project implements a real-time object detection system using the YOLOv5 (You Only

Look Once) architecture. The model is trained on the COCO dataset and tested for real-time inference

using OpenCV. The system can accurately detect multiple object classes such as persons, vehicles, and

animals in various environments.


RTRP Project Report: Object Detection Using Deep Learning

Introduction

Object detection combines object classification and object localization in a single framework. It has

applications in autonomous vehicles, surveillance, robotics, and augmented reality. This project

focuses on YOLOv5, known for its balance between speed and accuracy.

Objectives

- To understand object detection algorithms with a focus on YOLO.

- To implement a real-time object detection system using YOLOv5.

- To evaluate the model's performance using mAP (mean Average


Precision).

- To deploy the model using OpenCV for live camera input.

Literature Survey

"You Only Look Once: Unified, Real-Time Object Detection" by Redmon et al. (2016): Introduced YOLO

for real-time detection.

"YOLOv5: A PyTorch Implementation" by Ultralytics: Open-source YOLOv5 implementation with

enhanced speed and accuracy.

Methodology

Dataset Used:

- COCO dataset (Common Objects in


Context)

- 80 object classes

- Resized to 640x640

Model
Architecture:
RTRP Project Report: Object Detection Using Deep Learning

- YOLOv5s (small version)

- Backbone: CSPDarknet53

- Head: YOLO Layer for bounding box prediction

Training Configuration:

- Epochs: 100

- Batch Size: 16

- Optimizer: SGD

- Loss Function: Cross Entropy + GIoU


Loss

Tools & Technologies:

- Python, PyTorch, OpenCV, Google


Colab

Results & Evaluation

Metric | Value

[email protected] | 85.2%

Inference Speed | 25

FPS Model Size | ~14

MB

Live detection tested with webcam input. Detection shown with bounding boxes and
class labels.

Screenshots

Include images of detection results, training graphs, and webcam


inference.
RTRP Project Report: Object Detection Using Deep Learning

Conclusion

The project successfully demonstrated real-time object detection using YOLOv5. The system is capable

of accurately identifying multiple objects in live video feeds with minimal latency. Future improvements

include model quantization for mobile deployment and integration with edge devices like Jetson Nano.

Future Scope

- Deploy on mobile devices using TensorFlow Lite

- Integrate with robotics for obstacle avoidance

- Use custom datasets for specific industrial


applications

References

1. Redmon, J., et al. (2016). 'You Only Look Once: Unified, Real-Time Object
Detection.'

2. Ultralytics YOLOv5 GitHub Repo: https://github.com/ultralytics/yolov5

You might also like