ITAWD Computer Vision Project

Existing road image datasets frequently focus on European roads, which often exhibit limited variation in traffic and road conditions. To address this limitation, we developed the Indian Traffic and Adverse Weather Dataset (ITAWD), specifically designed to capture the unique characteristics of Indian roads. This is especially challenging due to the higher risk of collisions from moving objects and the lack of comprehensive datasets suited to these environments. The ITAWD features a total of 24,236 images categorized into 47 classes, including traffic objects, traffic lights, traffic signs, potholes, and cracks.

The dataset was created using two primary methods:

1. Dataset Creation from Real-World Scenes: We collected data from a range of driving sequences covering approximately 30 kilometers, recorded during both day and night across different locations in Mumbai. From this data, we selected around 8 sequences for inclusion in our dataset.

2. Datasets Merging : We integrated several existing datasets, including the Indian Driving Dataset (IDD)[1], the Road Damage Detection Dataset (RDD) [2], Indian Traffic Sign Dataset [3], BoschSmall Traffic Lights Dataset(BSTLD)[4] and Adverse Weather Detection Dataset [5] . Additionally, we used Roboflow \cite{roboflow} for re-annotation to address any missing labels.

[1] Girish Varma, Anbumani Subramanian, Anoop Namboodiri, Manmohan Chandraker, and CV Jawahar. Idd: A dataset for exploring problems of autonomous navigation in unconstrained environments. pages 1743–1751, 2019. [2] Deeksha Arya, Hiroya Maeda, Sanjay Kumar Ghosh, Durga Toshniwal, and Yoshihide Sekimoto. Rdd2020: An annotated image dataset for automatic road damage detection using deep learning. Data in Brief, 36:107133, 2021. [3] Rajesh Kannan Megalingam, Kondareddy Thanigundala, Sreevatsava Reddy Musani, Hemanth Nidamanuru, and Lokesh Gadde. Indian traffic sign detection and recognition using deep learning. International Journal of Transportation Science and Technology, 12(3):683–699, 2023. [4] Karsten Behrendt and Libor Novak. A deep learning approach to traffic lights: Detection, tracking, and classification. In Robotics and Automation (ICRA), 2017 IEEE International Conference on. IEEE. [5] Weatherdetection. Adverse weather detection dataset. https://universe.roboflow.com/weatherdetection/adverse-weather-detection, jul 2023.

ITAWD dataset was originally created by Shilpa Jain.