Research Article
Lightweight Object Detection Ensemble Framework for
Autonomous Vehicles in Challenging Weather Conditions
Rahee Walambe ,
1,2
Aboli Marathe ,
2
Ketan Kotecha ,
1,2
and George Ghinea
3
1
Symbiosis Institute of Technology, Symbiosis International University, Pune, India
2
Symbiosis Centre for Applied Artificial Intelligence, Symbiosis International University, Pune, India
3
Brunel University, London, UK
Correspondence should be addressed to Ketan Kotecha; drketankotecha@gmail.com
Received 22 July 2021; Revised 26 August 2021; Accepted 24 September 2021; Published 7 October 2021
Academic Editor: Anastasios D. Doulamis
Copyright © 2021 Rahee Walambe et al. is is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is
properly cited.
ecomputervisionsystemsdrivingautonomousvehiclesarejudgedbytheirabilitytodetectobjectsandobstaclesinthevicinity
of the vehicle in diverse environments. Enhancing this ability of a self-driving car to distinguish between the elements of its
environmentunderadverseconditionsisanimportantchallengeincomputervision.Forexample,poorweatherconditionslike
fog and rain lead to image corruption which can cause a drastic drop in object detection (OD) performance. e primary
navigationofautonomousvehiclesdependsontheeffectivenessoftheimageprocessingtechniquesappliedtothedatacollected
fromvariousvisualsensors.erefore,itisessentialtodevelopthecapabilitytodetectobjectslikevehiclesandpedestriansunder
challengingconditionssuchaslikeunpleasantweather.Ensemblingmultiplebaselinedeeplearningmodelsunderdifferentvoting
strategiesforobjectdetectionandutilizingdataaugmentationtoboostthemodels’performanceisproposedtosolvethisproblem.
e data augmentation technique is particularly useful and works with limited training data for OD applications. Furthermore,
using the baseline models significantly speeds up the OD process as compared to the custom models due to transfer learning.
erefore,theensemblingapproachcanbehighlyeffectiveinresource-constraineddevicesdeployedforautonomousvehiclesin
uncertain weather conditions. e applied techniques demonstrated an increase in accuracy over the baseline models and were
able to identify objects from the images captured in the adverse foggy and rainy weather conditions. e applied techniques
demonstrated an increase in accuracy over the baseline models and reached 32.75% mean average precision (mAP) and 52.56%
averageprecision(AP)indetectingcarsintheadversefogandrainweatherconditionspresentinthedataset.eeffectivenessof
multiple voting strategies for bounding box predictions on the dataset is also demonstrated. ese strategies help increase the
explainability of object detection in autonomous systems and improve the performance of the ensemble techniques over the
baseline models.
1. Introduction
e field of object detection (OD) has evolved from the
conceptualization of innovative algorithms to becoming an
integralpartofapplicationsintheindustry.eadoptionof
object detection in countless real-life applications has been
made possible due to the advancement of detection algo-
rithms and the increasing computational capabilities of
processors. From surveillance systems to scene under-
standing and face detection, object detection is being lev-
eragedtoassisthumansthroughintelligentanalyticsandby
automating arduous tasks. A recent application of object
detectionthatgarnersinterestisautonomousvehiclesdueto
the need for fast and accurate detectors for navigation
through traffic and urban environments.
In recent years, the rapid advancement of self-driving
carshastransformedtheirimagefromfuturisticvehiclesfar
ahead of our time to a part of an imaginable reality. e
diversity of features boasted by these vehicles is increasing
day by day, with special emphasis on the interpretability of
the car’s decisions, ethical considerations, and overall safety
[1,2].Designedusingmultiplelevelsofautomation,theself-
Hindawi
Computational Intelligence and Neuroscience
Volume 2021, Article ID 5278820, 12 pages
https://doi.org/10.1155/2021/5278820