Thanks to visit codestin.com
Credit goes to www.scribd.com

Scribd
Upload
3 views12 pages

Only Flags Final

The document outlines the development of an AI-based EO/IR sensor video classification system aimed at improving surveillance in low-visibility conditions. It details the system's capabilities to process thermal video, detect and classify objects like humans and vehicles, and identify movement in real-time using YOLOv8. The project utilizes a dataset from FLIR Systems and employs various technologies including Python, OpenCV, and PyTorch for implementation.

Uploaded by

wht10089
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
3 views12 pages

Only Flags Final

The document outlines the development of an AI-based EO/IR sensor video classification system aimed at improving surveillance in low-visibility conditions. It details the system's capabilities to process thermal video, detect and classify objects like humans and vehicles, and identify movement in real-time using YOLOv8. The project utilizes a dataset from FLIR Systems and employs various technologies including Python, OpenCV, and PyTorch for implementation.

Uploaded by

wht10089
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/ 12

0NLY_FL4G$

EO/IR Sensor Video

Image Classification

& Identification

System
OUR TEAM

Yash Saraogi Tallam Sri Sai Yash Gautam Sarthak Sharma

AI/ML Developer CV Developer Model Evaluation & Testing


Subramanyam
Engineer
Backend Developer
Problem Statement

Traditional surveillance systems are ineffective in


low-visibility conditions.
EO/IR sensors capture thermal imagery useful for
day/night object tracking.
Need a system that:
Processes EO/IR video input
Detects and classifies multiple objects (like
humans, vehicles)
Identifies static vs. moving objects
Works in real-time or near-real-time
Objective

Build an end-to-end AI system that:


Accepts EO/IR (thermal) video as input
Performs object detection using YOLOv8
Classifies objects like humans, vehicles, etc.
Identifies movement by comparing consecutive frames
Visualizes the results with bounding boxes and labels
System

Architecture

Each IR video frame


is analyzed using a
pre-trained AI
model. Detected
objects are labeled.
Motion is estimated
by comparing frame
differences.
Dataset - FLIR

Thermal Dataset

Dataset: FLIR ADAS Thermal Dataset


Source: FLIR Systems Inc. (free for academic use)
Images: 14,452 thermal images with bounding box
annotations
Classes: Person, Car, Bicycle
Format: COCO-style JSON annotations
Use: Training YOLOv8 for object detection on thermal
images
TOOLS AND TECHNOLOGY

STACK
This project will be developed using Python for its strong
support in AI and computer vision. We used YOLOv8 for object
detection, as it offers high speed and accuracy, especially on
image-based tasks. OpenCV was used for handling video input,
frame processing, and drawing detection results. PyTorch
powered our deep learning model integration. For visualizing
results and creating a simple user interface, we used Streamlit.
Dataset annotation and model preparation were supported by
tools like Roboflow.
Sample Output &

Results

System detects people and vehicles in IR footage


Able to label objects with class names
Optional: Movement shown with colored boxes
Can process live feed or stored video
Challenges

X
Frame rate drop when
running GPU
X? Limited number of
classes in IR datasets
X Low contrast in
thermal images
Conclusion

Developed a thermal video classification system using AI


Detects and classifies humans, vehicles from IR data
Can distinguish moving/static objects
Useful for surveillance, rescue, military and UAV use-cases
Scalable, extendable, and deployable

“Thermal vision + AI: unlocking human-like perception for


machines.”
REFERENCES
REFERENCES

1. YOLOv8 by Ultralytics: https://docs.ultralytics.com/


2. OpenCV Documentation: https://docs.opencv.org/
3. Object tracking: https://github.com/abewley/sort
Thank you!!

You might also like