Kaggle Dataset Information

Dataset Details

Dataset Name/URL Slug: dhaka-traffic-classification-4-levels Dataset Title: "Dhaka City Traffic Classification Dataset - 4-Level Congestion Analysis"

Dhaka City Traffic Classification Dataset - 4-Level Congestion Analysis

Overview

This dataset is a refined version of the original DhakaAI traffic detection dataset, specifically preprocessed and categorized for 4-level traffic congestion classification in Dhaka City, Bangladesh. The dataset has been organized to support machine learning research in urban traffic analysis and intelligent transportation systems.

Dataset Origin

Source: Based on the DhakaAI Dhaka-based Traffic Detection Dataset Original Dataset: https://www.kaggle.com/datasets/rifat963/dhakaai-dhaka-based-traffic-detection-dataset Preprocessing: Images have been categorized into 4 distinct traffic levels for classification tasks

Traffic Categories

1. No Traffic (Class 0)

• Description: Free-flowing roads with minimal to no vehicles
• Characteristics: Clear roadways, unobstructed movement
• Use Case: Baseline for traffic-free conditions

2. Light Traffic (Class 1)

• Description: Minimal congestion with smooth vehicle flow
• Characteristics: Few vehicles present, normal speed movement
• Use Case: Optimal traffic conditions

3. Moderate Traffic (Class 2)

• Description: Some congestion present, reduced but steady flow
• Characteristics: Noticeable vehicle density, slower movement
• Use Case: Warning level for traffic management

4. Heavy Traffic (Class 3)

• Description: Significant congestion with restricted movement
• Characteristics: High vehicle density, stop-and-go patterns
• Use Case: Critical level requiring traffic intervention

Dataset Structure

dataset/
├── Train/
│  ├── no traffic/
│  ├── light traffic/
│  ├── moderate traffic/
│  └── heavy traffic/
└── Test/
  ├── no traffic/
  ├── light traffic/
  ├── moderate traffic/
  └── heavy traffic/

Technical Specifications

• Image Format: JPEG/PNG
• Color Space: RGB
• Recommended Input Sizes: 128x128, 256x256, 512x512 pixels
• Classes: 4 (balanced distribution recommended)
• Split: Pre-divided into training and testing sets

Applications

• Traffic congestion prediction
• Intelligent transportation systems
• Urban planning analysis
• Real-time traffic monitoring
• Deep learning model benchmarking

Model Performance Benchmarks

This dataset has been tested with multiple deep learning architectures: - EfficientNetB0: 54.17% accuracy (best performance) - MobileNetV2: 45.50% accuracy - Custom CNN: 35.83% accuracy - ResNet50: 33.17% accuracy

Usage Examples

Perfect for: - Computer vision research in traffic analysis - Comparative studies of CNN architectures - Urban traffic pattern recognition - Transportation engineering projects - Academic research in machine learning

Citation

If you use this dataset in your research, please cite: bibtex @dataset{dhaka_traffic_4levels_2025, title={Dhaka City Traffic Classification Dataset - 4-Level Congestion Analysis}, author={Md. Roman Bin Jalal}, year={2025}, publisher={Kaggle}, url={https://www.kaggle.com/datasets/mdromanbinjalal/dhaka-traffic-classification-4-levels}, note={Derived from DhakaAI Traffic Detection Dataset by rifat963} }

Please also cite the original dataset: bibtex @dataset{dhakaai_original_2023, title={DhakaAI Dhaka-based Traffic Detection Dataset}, author={rifat963}, year={2023}, publisher={Kaggle}, url={https://www.kaggle.com/datasets/rifat963/dhakaai-dhaka-based-traffic-detection-dataset} }

License

License

Please respect the original dataset's license terms and provide appropriate attribution.

Acknowledgments

• Original dataset created by rifat963
• Dhaka city traffic data collection
• Traffic categorization and preprocessing for ML applications

Dataset

This dataset was created by Harikrishnan Vamsi

Contents

Udacity Self Driving Car

Overview

The original Udacity Self Driving Car Dataset is missing labels for thousands of pedestrians, bikers, cars, and traffic lights. This will result in poor model performance. When used in the context of self driving cars, this could even lead to human fatalities.

Re-labeled the dataset to correct errors and omissions.

Content

The dataset contains 97,942 labels across 11 classes and 15,000 images. There are 1,720 null examples (images with no labels).

All images are 1920x1200 (download size ~3.1 GB).

Annotations have been hand-checked for accuracy by Roboflow.

https://i.imgur.com/bOFkueI.pnghttps://" alt="Class Balance">

Annotation Distribution: https://i.imgur.com/NwcrQKK.png" alt="Annotation Heatmap">

Use Cases

Udacity is building an open source self driving car! You might also try using this dataset to do person-detection and tracking.

Note: the dataset contains many duplicated bounding boxes for the same subject which we have not corrected. You will probably want to filter them by taking the IOU for classes that are 100% overlapping or it could affect your model performance (expecially in stoplight detection which seems to suffer from an especially severe case of duplicated bounding boxes).

About Roboflow

Roboflow makes managing, preprocessing, augmenting, and versioning datasets for computer vision seamless.

Developers reduce 50% of their boilerplate code when using Roboflow's workflow, save training time, and increase model reproducibility. :fa-spacer:

Acknowledgements

Licensed by MIT. More details in the README.txt files. Provided by Roboflow License: MIT