NITCAD — An object detection, classification and stereo vision dataset for autonomous navigation in Indian roads

The area of autonomous vehicles (AVs) is booming. One particular domain which helps in the research of AVs is labelled data because it helps the ML algorithms (in most cases CNNs) to understand the scenario and act according to it.

India has some unique scenarios when it comes to roads such as

1. Auto rickshaws are ubiquitous (which are not common in foreign roads, thus not available in foreign datasets)
2. Unstructured environment in many roads
3. Lack of lanes/dividers

An example image from NITCAD dataset where you can observe the lack of lanes and dividers.

So, it is important to collect, label and opensource data. As a small contribution, we have created an annotated dataset and published it in a conference. Please refer to https://doi.org/10.1016/j.procs.2020.04.022 for full paper which includes the evaluation of dataset and comparison of various metrics. This blog helps as an overview of our work and access dataset (fill the Google Form attached at the bottom).

NITCAD — An object detection, classification and stereo vision dataset for autonomous navigation in Indian roads by Namburi GNVV Satya Sai Srinath, Athul Zac Joseph, S Umamaheswaran, Ch. Lakshmi Priyanka, Malavika Nair M, Praveen Sankaran is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License. Based on a work at https://www.sciencedirect.com/science/article/pii/S187705092030987X?via%3Dihub%3C/a%3E

Note: This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Dataset Specifications

Entire NITCAD dataset has been divided into 2 parts

1. NITCAD object dataset : Collected by using Noise Play 2 Action camera in 720p at 30fps.
2. NITCAD stereo vision dataset : Collected by using Intel Realsense Depth camera D435

Data is collected in various rural and urban areas in Kerala state.

There are about 11,000 images with 4800 images manually labelled using ‘LabelBox’

There are 7 classes in NITCAD object dataset. From left to right: Car, Bus, Pedestrian, Two wheeler, Truck, Van and Auto rickshaw.

Some of the statistics of NITCAD object dataset. Left: Total number of object present per class. Right: Total number of images having a specific class. One conclusion is — Car, Autos and Two wheelers are most common.

An example image pair from NITCAD stereo vision dataset.

Please fill this Google form to get access to NITCAD dataset.

Github link: https://github.com/NamburiSrinath/NITCAD-dataset (This repository contains our entire major project in which NITCAD dataset is a part)

Please mail to namburisrinath@gmail.com, psankaran@nitc.ac.in in case you have further doubts regarding dataset/paper with mail subject as “NITCAD Dataset — Doubts”

Note: NITCAD Dataset stands for National Institute Of Technology Calicut Autonomous Driving dataset

P.S: I would like to thank my project teammates Athul Zac Joseph, Ch. Lakshmi Priyanka, Malavika Nair M, S Umamaheswaran and our guide Dr. Praveen Sankaran who helped at various stages during the project.

Personal Website: namburisrinath.github.io

Medium Handle: namburisrinath.medium.com

LinkedIn: https://www.linkedin.com/in/namburi-gnvv-satya-sai-srinath/