Swayansaasita is a research and development initiative in Autonomous Vehicles (AV) with focus on challanges in developing countries like India. The project also focuses on design and development of ADAS systems for 2-wheeler vehicles.
Download Dataset Cite this project.
The availability of high data is the key element for research and development of intelligent systems for autonomous driving. Such data also helps in designs and development of simulations and models for studying driving dynamics - from the drivers PoV and also helps in understanding the dynamics of other elements on the road during driving such as pedestrians and other vehicles.
Through this project, we are releasing the biggest available multi-modal dataset for research and development in autonomous driving and traffic behavior studies. The following key objectives are the focus of Swayansaasita.
Largest Multimodal Dataset
Largest multimodal dataset from real world including 360 degree point clouds, multiple IMU sensors, GPS.
Annotated 3D data for ML.
Annotated data including annotated 3D pointclouds, IMU Events and Images for ML applications.
Our software.
We develop simulation environment cum UI to interact with the dataset and run experiments.
We train predictive ML models.
We experiment with ML algorithms to devleop Advanced Drivers Assistens Systems.
Technical writeups on the technologies involved in this project.
Computer vision plays a crucial role in enabling self-driving abilities in motor vehicles. In this blog, we will explore how computer vision, along with other sensor data, such as multi-cameras, LiDAR, IMU, and GPS, can be utilized to collect data and make predictions for self-driving motorcycles.
Read blogComputer vision is a field of artificial intelligence that enables machines to interpret and understand visual information from the world. It involves acquiring, processing, and analyzing images and videos to recognize patterns, detect objects, and make decisions. Applications include facial recognition, autonomous vehicles, medical imaging, and industrial automation, enhancing how machines perceive and interact with their environments.
Read blogROS (Robot Operating System) is an open-source framework for developing robotic software. It provides tools, libraries, and conventions to simplify the creation of complex robot behaviors across various robotic platforms. ROS facilitates communication between components, hardware abstraction, and device control, allowing developers to focus on application-specific tasks. It supports modular design, making it widely used in research and industry for robotics development.
Read blogPoint clouds are collections of data points in three-dimensional space, typically generated by 3D scanning technologies like LiDAR or stereo cameras. Each point represents a sample on the surface of an object, capturing its shape and geometry. Point clouds are widely used in fields such as computer vision, robotics, and geospatial analysis for tasks like 3D modeling, object recognition, and environment mapping.
Read blogGNSS (Global Navigation Satellite System) is a network of satellites that provide geolocation and time information to GNSS receivers on Earth. It includes systems like GPS (USA), GLONASS (Russia), Galileo (EU), and BeiDou (China). GNSS enables accurate positioning and navigation for various applications, including mapping, surveying, transportation, and autonomous vehicles, by determining a receiver's exact location globally.
Computer vision plays a crucial role in enabling self-driving abilities in motor vehicles. In this blog, we will explore how computer vision, along with other sensor data, such as multi-cameras, LiDAR, IMU, and GPS, can be utilized to collect data and make predictions for self-driving motorcycles.
Read blogHardware-software integration in robotics involves combining physical components like sensors, actuators, and processors with software systems that control and coordinate them. This integration ensures seamless communication between hardware and software, enabling robots to perceive their environment, make decisions, and perform tasks effectively. It is crucial for achieving reliable, real-time performance in various robotic applications, from industrial automation to autonomous navigation.
Read blogMachine learning-based prediction involves using algorithms to analyze data and identify patterns to forecast future outcomes. By training on historical data, models learn to predict values, classifications, or trends in unseen data. Applications range from predicting stock prices and weather conditions to identifying equipment failures and personalizing content. These predictions improve decision-making and automate complex processes in various fields.
Read blogSwayansaasita: Phase-I is a research and development initiative by the Subhankar Mishra Lab, funded and supported by La Fondation Dassault Systèmes, India.
Dataset will be released soon.
We are currnetly in the process of collecting, post processing and annotating the dataset. The Data Collection plan is available to refer here.
BibTex citation for this project is attached below. The contents of this project are availabe for research purposes and can be used with due citation to this project.
@misc{jyothish_smishra_2024,
title = {Swayansaasita - Autonomous Two-Wheeler Driving - Phase 1},
author = {Jyothish, K.J., Keshri, S., Vishwakarma, A., and Mishra, S.},
year = {2024},
howpublished = {\url{https://www.niser.ac.in/~smishra/project/cs2303/}},
note = {Accessed: April 29, 2024}
}