Object Detection for Autonomous Driving: High-Dynamic Range vs. Low-Dynamic Range Images

Date

2022-01-01

Author

Kocdemir, Ismail H.
Akyüz, Ahmet Oğuz
Koz, Alper
Chalmers, Alan
Alatan, Abdullah Aydın
Kalkan, Sinan

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

144
views

0
downloads

An important problem in autonomous driving is to perceive objects even under challenging illumination conditions. Despite this problem, existing solutions use low-dynamic range (LDR) images for object detection for autonomous driving. In this paper, we provide a novel analysis on whether high-dynamic range (HDR) images can provide better performance for object detection for autonomous driving. To this end, we choose a seminal deep object detector and systematically evaluate its performance when trained with (i) LDR images, (ii) HDR images, and (iii) tone-mapped LDR images for scenes with different illuminations. We show that a detector with HDR images pre-processed with normalization and gamma correction can only marginally perform better than a detector with LDR or tone-mapped LDR images. Our analysis of this unexpected finding reveals that a detector with HDR images requires significantly more samples as the space of HDR images is significantly larger than that of LDR images.

Subject Keywords

Autonomous Driving, Object Detection, High-Dynamic Range (HDR), Low-Dynamic Range (LDR), Autonomous Driving, High-Dynamic Range (HDR), Low-Dynamic Range (LDR), Object Detection

URI

https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85143629419&origin=inward
https://hdl.handle.net/11511/101413

DOI

https://doi.org/10.1109/mmsp55362.2022.9949582

Conference Name

24th IEEE International Workshop on Multimedia Signal Processing, MMSP 2022

Collections

Department of Computer Engineering, Conference / Seminar

Suggestions

OpenMETU
Core

AN ANALYSIS ON THE EFFECT OF DYNAMIC RANGE ON OBJECT DETECTION WITH DEEP NEURAL NETWORKS Koçdemir, İsmail Hakkı; Kalkan, Sinan; Alatan, Abdullah Aydın; Department of Computer Engineering (2021-10-8) An important problem in computer vision, particularly in object detection, is being able to perceive objects even under challenging illumination conditions. Being robust to such conditions is especially important in applications, such as autonomous driving. Despite the significance of the problem, existing autonomous driving systems use deep object detection networks with low-dynamic range (LDR) images during both the training phase and the testing phase. In this thesis, we investigate whether high-dynamic ...
HDR IMAGE CONSTRUCTION FROM TRIFOCAL MULTIEXPOSURE IMAGES KOZ, ALPER; Demirkılıç, Barış; Kurt, Yunus Bilge; Akyüz, Ahmet Oğuz; Kalkan, Sinan; Alatan, Abdullah Aydın; Chalmers, Alan (2021-10-06) With the progress of autonomous vehicles, the sensing of the environment in more detail with higher dynamic ranges has become more important to classify surrounding objects and obstacles. While stereo HDR images for this purpose can provide advantages compared to the conventional LDR images, they suffer from limited dynamic ranges and spike-like noises due to the inaccuracies in disparity estimation. In this paper, we formulate the HDR image construction problem from trifocal multi-exposure images and devel...
Feedback motion planning of unmanned surface vehicles via random sequential composition Ege, Emre; Ankaralı, Mustafa Mert (SAGE Publications, 2019-08-01) In this paper, we propose a new motion planning method that aims to robustly and computationally efficiently solve path planning and navigation problems for unmanned surface vehicles (USVs). Our approach is based on synthesizing two different existing methodologies: sequential composition of dynamic behaviours and rapidly exploring random trees (RRT). The main motivation of this integrated solution is to develop a robust feedback-based and yet computationally feasible motion planning algorithm for USVs. In ...
Evaluation of deep learning based multiple object trackers Moured, Omar; Akar, Gözde; Department of Electrical and Electronics Engineering (2020-9) Multiple object tracking (MOT) is a significant problem in the computer vision community due to its applications, including but not limited to, surveillance and emerging autonomous vehicles. The difficulties of this problem lie in several challenges, such as frequent occlusion, interaction, intra-class variations, in-and-out objects, etc. Recently, deep learning MOT methods confront these challenges effectively. State-of-the-art deep learning (DL) trackers pipeline consists of two stages, i.e., appearance h...
Mini autonomous car architecture for urban driving scenarios Karabulut, Gökhan; Can, Tolga; Department of Computer Engineering (2019) Autonomous cars capable of driving in city traffic have been long studied in architectures decomposed into perception, planning, and control components. Recent advances in deep learning techniques considerably contributed to the perception component of this approach. These techniques also laid the groundwork for the progress of other approaches such as end-to-end learning of steering commands and driving affordances from camera images. Though these approaches are promising to simplify the overall architectu...

Citation Formats

I. H. Kocdemir, A. O. Akyüz, A. Koz, A. Chalmers, A. A. Alatan, and S. Kalkan, “Object Detection for Autonomous Driving: High-Dynamic Range vs. Low-Dynamic Range Images,” presented at the 24th IEEE International Workshop on Multimedia Signal Processing, MMSP 2022, Shanghai, Çin, 2022, Accessed: 00, 2023. [Online]. Available: https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85143629419&origin=inward.