Real-Time Object Detection using
YOLOv8
• A Fast and Accurate Deep Learning Approach
• Presented by: Your Name
• Course/Semester: B.Voc AI & ML – 4th
Semester
• Date: DD/MM/YYYY
Introduction
• - Object detection identifies and locates
objects in images or videos.
• - Real-time detection is critical for safety,
automation, and smart systems.
• - YOLOv8 is the latest and fastest version of
the YOLO family.
Problem Statement
• - Traditional detection models are slow or less
accurate.
• - Need for:
• • High accuracy
• • Low latency real-time detection
• • Scalability for multiple applications
Objectives
• - Implement YOLOv8 for real-time detection.
• - Train or fine-tune model with custom
datasets.
• - Achieve high FPS and better accuracy for
practical deployment.
Methodology
• Start → Data Collection & Labeling → Model
Training → Real-Time Inference → Evaluation
→ Deployment → Feedback → End
Tools & Technologies
• - Programming: Python
• - Libraries: Ultralytics YOLOv8, OpenCV,
PyTorch
• - Hardware: GPU-enabled system,
Webcam/Camera
• - Optional: TensorRT or ONNX for faster
deployment
Dataset
• - Dataset from COCO or custom-labeled
images (Roboflow, LabelImg).
• - Split: 70% Training, 20% Validation, 10%
Testing.
• - Classes: Person, Car, Bike, Dog, etc.
Model Training (Code)
• from ultralytics import YOLO
• model = YOLO('yolov8n.pt')
• model.train(data='data.yaml', epochs=50,
imgsz=640, batch=16)
• model.val()
Real-Time Inference (Code)
• from ultralytics import YOLO
• import cv2, time
• model =
YOLO('runs/detect/train/weights/best.pt')
• cap = cv2.VideoCapture(0)
• while True:
• ret, frame = cap.read()
• results = model(frame)
Real-Time Output
• - Screenshot examples of detections with
labels.
• - FPS displayed on screen for performance.
• - Include a short demo clip if possible.
Results
• - Accuracy: mAP@0.5, Precision, Recall
• - Speed: FPS achieved during live testing
• - Sample Output: Objects detected with
bounding boxes
Optimization
• - Use smaller models (yolov8n.pt) for speed.
• - Run on GPU for better performance.
• - Export models to ONNX or TensorRT for
deployment.
Applications
• - Smart Surveillance – Intruder detection.
• - Traffic Monitoring – Vehicle counting,
accident alerts.
• - Robotics – Navigation and object interaction.
• - Retail Analytics – Customer behavior
tracking.
Future Scope
• - Edge device deployment (Jetson Nano,
Raspberry Pi).
• - Integrate with tracking algorithms like
ByteTrack.
• - Multi-camera and cloud-based systems.
Conclusion
• - YOLOv8 enables fast and accurate real-time
object detection.
• - Suitable for multiple industries and scalable
applications.
• - Future improvements will make detection
even more efficient.
References
• - Ultralytics YOLOv8 Documentation:
https://docs.ultralytics.com
• - COCO Dataset: https://cocodataset.org
• - Tutorials and GitHub resources
