Pakistan Traffic Signs Dataset

Pakistani Traffic Sign Dataset

Description

The Pakistani Traffic Sign Dataset is a real-world image collection capturing traffic signs from across Pakistan. The dataset is designed to support computer vision research and the development of AI-powered transportation systems tailored to Pakistan's unique road signage landscape, which features signs in both Urdu and English scripts.

Data Collection

The dataset consists of high-resolution images of traffic signs collected from diverse urban and rural regions across Pakistan, capturing real-world conditions with variations in camera devices, distances, angles, and environmental settings. The collection reflects the authentic signage landscape of Pakistani roads, including both Urdu and English language signs.

Preprocessing

Images have undergone standard preprocessing steps including resizing and normalization. Some images intentionally retain natural noise, blur, and occlusions to preserve real-world conditions and improve model generalization in practical deployment scenarios.

Unique Aspect

A distinguishing feature of this dataset is the inclusion of traffic signs in both Urdu and English, reflecting Pakistan's bilingual road signage system. This makes the dataset particularly valuable for multilingual sign recognition, Urdu OCR, and cross-lingual traffic analysis tasks.

Environmental Diversity

Images are captured under a broad range of environmental and lighting conditions including daytime, nighttime, shadows, varying weather, and complex urban and rural backgrounds, enhancing model robustness across diverse real-world scenarios.

Use Cases

This dataset is well-suited for a range of computer vision and AI tasks, including:

• Traffic sign detection and classification
• Urdu and English multilingual OCR on road signage
• Object detection and localization
• Training AI models for intelligent transportation systems (ITS)
• Autonomous driving research in South Asian road environments

Limitations

• Certain traffic sign categories may be underrepresented, leading to class imbalance across the dataset.
• Some images may contain occlusions, physical damage, or weathering that affects sign visibility.
• Users are advised to apply data augmentation or class-balancing techniques depending on their specific use case.

Sample Data

{
  "images": [
    { "id": 1,  "file_name": "images/0674418c-7d8c-41b8-832b-2583b3654f5a.jpg", "width": 4536, "height": 8064 },
    { "id": 2,  "file_name": "images/002fa57f-9f0f-475e-9143-abce8a1c9150.jpg", "width": 4536, "height": 8064 },
    { "id": 3,  "file_name": "images/9300ae95-2e41-4c7b-a5bd-cc154274f115.jpg", "width": 2268, "height": 4032 },
    { "id": 4,  "file_name": "images/18358ac1-feed-48ce-90eb-1743b293fe39.jpg", "width": 4536, "height": 8064 },
    { "id": 5,  "file_name": "images/a702d265-7cec-41f1-a919-3ae263fcaa8b.jpg", "width": 4536, "height": 8064 },