DATS_2022: A versatile indian dataset for object detection in unstructured traffic conditions

Area of Science:

• Computer Vision
• Machine Learning
• Artificial Intelligence

Background:

• Driver Assistance Systems (DAS) are crucial for navigating India's complex traffic.
• Developing accurate machine learning algorithms for DAS requires specialized datasets.
• Existing datasets often lack the diversity and detail needed for Indian traffic conditions.

Purpose of the Study:

• To introduce DATS_2022, a novel dataset for object detection tailored to Indian traffic scenarios.
• To provide a comprehensive resource for training and evaluating machine learning models for DAS.
• To support advancements in visual scene understanding for autonomous and semi-autonomous vehicles.

Main Methods:

• Collected over 10,000 high-resolution images from diverse Indian rural and urban traffic environments using Android phones.
• Annotated images using a free, open-source tool, generating XML files for machine learning.
• Included more than 7,000 annotations across 45 distinct object classes.

Main Results:

• The DATS_2022 dataset features a wide array of objects relevant to Indian traffic.
• The dataset encompasses both rural and urban driving conditions, offering broad applicability.
• Annotations are provided in multiple formats, facilitating diverse machine learning approaches.

Conclusions:

• DATS_2022 is a valuable resource for researchers in object detection and classification.
• The dataset will accelerate the development of more robust and accurate driver assistance systems for India.
• Facilitates deep learning research for improved traffic scene understanding and vehicle safety.