PDrive20k serves as a pivotal tool for both researchers and companies in the autonomous driving sector, highlighting the distinct challenges and prospects presented by Pakistan's roadways. From delving into urban traffic complexities to maneuvering rugged landscapes or navigating the lively lanes of Pakistani cities, this dataset offers an in-depth reservoir of data. This aids in crafting autonomous driving solutions specifically for this unique region. Dive in with us, as we uncover the essence of Pakistan's roads and propel innovation.

A scenic view of Pakistan Road

Diversity in PDrive20k Data

The autonomous driving dataset from Pakistan is a treasure trove of diverse and unique data, offering a rich tapestry of scenes and conditions that reflect the intricacies of Pakistan's roadways. From the bustling streets of major cities to the serene landscapes of rural areas, this dataset encapsulates the essence of Pakistan's road environments. Here's a glimpse into the remarkable diversity it encompasses:

Everything YOu Need

3D Object Bounding Boxes

Our dataset includes detailed annotations of 3D object bounding boxes. This information is crucial for object detection and tracking in autonomous driving scenarios. Researchers and developers can use these annotations to train and evaluate object detection algorithms, ensuring the accurate perception of objects in the environment.

Lane Lines

(Segmentation + XY Coordinates)

Accurate lane detection is vital for autonomous vehicles to navigate safely on roads. In our dataset, you'll find lane line annotations that include both segmentation masks and XY coordinates. This allows for the precise identification and localization of lane boundaries, aiding in the development of lane-keeping and path planning algorithms.


(Vehicle, Pedestrian, Road)

Semantic segmentation is a fundamental task for autonomous vehicles to understand their surroundings. Our dataset provides high-quality segmentation masks for various classes, including vehicles, pedestrians, and road surfaces. This segmentation information aids in object recognition, obstacle avoidance, and road scene understanding, essential for the safe operation of autonomous vehicles.

What Makes It Different?

24 Unique Sequences:

PDrive20K consist of 20,000 labeled images with 24 distinct sequences, this dataset provides a comprehensive view of Pakistan's roads in various scenarios and conditions.

Pedestrians in Traditional Clothing:

Pakistan's rich cultural diversity is on full display, with pedestrians donning traditional attire like shalwar kameez and burkha. Recognizing and responding to individuals in traditional clothing is crucial for the safety of autonomous vehicles, and this dataset offers ample opportunities for training such recognition systems.

Day and Night Data:

The dataset captures the stark contrast between day and night driving conditions. With scenes ranging from the vibrant hues of daylight to the challenges of low-light and nighttime driving, researchers and autonomous driving companies gain insights into adapting their algorithms and sensors for different lighting conditions.

By offering all these elements within a single dataset, we aim to simplify and streamline the development and testing of autonomous driving systems. Whether you're working on object detection, lane following, or semantic segmentation, our dataset provides the diverse and comprehensive data you need to drive innovation in the field of autonomous mobility.

Uncommon Construction Vehicles:

Pakistan's construction industry boasts a fascinating array of vehicles rarely seen in other environments. From distinctive bulldozers to specialized excavation equipment, this dataset provides a unique opportunity to train autonomous systems to recognize and navigate around these unconventional vehicles.

Auto Rickshaws:

A common sight on Pakistani roads, auto rickshaws present a unique challenge due to their nimble nature and frequent stops. The dataset includes data on these three-wheeled vehicles, allowing developers to enhance the safety and efficiency of autonomous vehicles around them.

Trucks Adorned with Traditional Truck Art:

Pakistan's iconic truck art adorns countless commercial vehicles, transforming them into vibrant and intricately decorated masterpieces. This dataset features these ornate trucks, enabling developers to build models that can identify and interact with them, further enriching the autonomous driving experience in Pakistan.

PDrive20K Download

Download the dataset now and reshape your perspective. Or, begin with our starter guide notebook to ease into utilizing the dataset. Share your innovations using #PDrive20k. Best of luck!

Download Dataset Starter Guide