finetune_dataset_info.md

Fine-tuning Dataset Information

Overview

This directory contains the fine-tuning dataset used to improve the YOLOv8m model's performance on truck and bus detection.

Dataset Statistics

• Total Images: 92
• Total Objects: 2,277
• Classes: 8 vehicle types

Class Distribution

• bus: 57 objects
• car: 1,900 objects
• lcv: 188 objects
• multiaxle: 10 objects
• truck: 121 objects

File Structure

finetune_dataset/
├── images/           # 92 training images
├── labels/           # 92 corresponding label files
├── README.dataset.txt # Dataset metadata
└── README.roboflow.txt # Roboflow export information

Usage

This dataset was used for Stage 2 fine-tuning to address the truck/motorcycle misclassification issue found in Stage 1 training.

Format

• Images: JPG format
• Labels: YOLO format (.txt files)
• Annotations: Bounding boxes with class IDs

Quality

All images are high-quality highway traffic scenes with clear vehicle visibility and proper annotations.