MANOHAR N

Dr. MANOHAR N

orcid

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
20 January 2025 in MethodsX
Verified
Journal Article
Open access
HYH. S. Gowri Yaamini
H. S. Gowri Yaamini
SJSwathi K J
Swathi K J
MNMANOHAR N
MANOHAR N
AGAjay Kumar G
Ajay Kumar G
Accurate and precise detection of lanes and traffic signs is predominant for the safety and efficiency of autonomous vehicles and these two significant tasks should be addressed to handle Indian traffic conditions. There are several state-of-art You Only Live Once (YOLO) models trained on benchmark datasets which fails to cater the challenges of Indian roads. To address these issues, the models need to be trained with a wide variety of Indian data samples for the autonomous vehicles to perform better in India. YOLOv8 algorithm has its challenges but gives better precision results and YOLOv8 nano variant is widely used as it is computationally less complex comparatively. Through rigorous evaluations of diverseness in the datasets, the proposed YOLOv8n transfer learning models exhibits remarkable performance with a mean Average Precision (mAP) of 90.6 % and inference speed of 117 frames per second (fps) for lane detection whereas, a notable mAP of 81.3 % for traffic sign detection model with a processing speed of 56 fps.•YOLOv8n Transfer Learning approach by adjusting architecture for lane and traffic sign detection in Indian diverse Urban, Suburban, and Highway scenarios.•Dataset with 22,400 images of normal and complex Indian scenarios include crude weathering of roads, traffic conditions, diverse tropical weather conditions, partially occluded and partially erased lanes, and traffic signs.•The model performance with notable precision and frame wise inference.

1 February 2022 in International Journal of Electrical and Computer Engineering (IJECE)
Verified
Journal Article
Open access
PDF file
SRShobha Rani
Shobha Rani
MNMANOHAR N
MANOHAR N
HMHariprasad M.
Hariprasad M.
PRPushpa B R
Pushpa B R
Automated reading of handwritten Kannada documents is highly challenging due to the presence of vowels, consonants and its modifiers. The variable nature of handwriting styles aggravates the complexity of machine based reading of handwritten vowels and consonants. In this paper, our investigation is inclined towards design of a deep convolution network with capsule and routing layers to efficiently recognize Kannada handwritten characters. Capsule network architecture is built of an input layer, two convolution layers, primary capsule, routing capsule layers followed by tri-level dense convolution layer and an output layer. For experimentation, datasets are collected from more than 100 users for creation of training data samples of about 7769 comprising of 49 classes. Test samples of all the 49 classes are again collected separately from 3 to 5 users creating a total of 245 samples for novel patterns. It is inferred from performance evaluation; a loss of 0.66% is obtained in the classification process and for 43 classes precision of 100% is achieved with an accuracy of 99%. An average accuracy of 95% is achieved for all remaining 6 classes with an average precision of 89%.

11 December 2024 in IEEE Access
Verified
Journal Article
Open access
SRSuresha R
Suresha R
MNMANOHAR N
MANOHAR N
GKGURUMADAIAH AJAY KUMAR
GURUMADAIAH AJAY KUMAR
MSM. Rohit Singh
M. Rohit Singh
Detecting small traffic signs is essential for autonomous vehicles (AV) to operate safely, adhere to traffic regulations, navigate effectively, maintain situational awareness, adapt to local conditions, and overcome the technical challenges associated with diverse environments. As AV technology continues to evolve, improving the accuracy and reliability of small sign detection remains a crucial area of research and development. Enforcing small traffic sign detection systems (TSDS) is essential for road safety, ensuring drivers know important information, warnings, and regulations aligned with smart cities, enhancing traffic management, and optimizing signal timings. Over the past ten years, many deep-learning techniques have been reported for TSDS to identify tiny traffic regions. This review comprehensively examines the performance of state-of-the-art deep learning models, including YOLO (You Only Look Once), SSD (Single Shot MultiBox Detector), and various RCNN (Region-based Convolutional Neural Network) variants, assessing their strengths and weaknesses for small traffic sign detection through detailed tables and bar graphs. Additionally, the review examines key evaluation metrics used for TSDS and explores standard benchmark datasets like TT100k, GTSDB, CCTSD, STS, and DFG. It emphasizes the dataset’s attributes and complex factors affecting detection performance. The analysis focuses on how various deep learning models perform on these standard datasets, presenting the results in tables and comparative bar graphs. Finally, the review discusses current challenges in TSDS technology and proposes recommendations for future development of driver-aid systems. This comprehensive analysis aims to guide the design of future computer vision systems for small traffic sign detection.

8 April 2023
Verified
Book Chapter
Closed access
PDF file
VJV. Jhanavi
V. Jhanavi
ANA. Sanjay Nag
A. Sanjay Nag
MNMANOHAR N
MANOHAR N

6 February 2026 in IEEE Access
Verified
Journal Article
Open access
PMPriyanka M
Priyanka M
MNMANOHAR N
MANOHAR N
SRSuresha R
Suresha R
RDR. Dayananda
R. Dayananda
SNSmitha Sunil Kumaran Nair
Smitha Sunil Kumaran Nair
TST. Sheela
T. Sheela
Bone Age Estimation is crucial for evaluating skeletal maturation, particularly in identifying differences between a minor’s chronological and bone age. It contributes to the development of robust institutions by ensuring that individuals, especially children, are treated fairly in accordance with their actual ages within their respective jurisdictions. The examination is subjective, as the radiologist evaluates each bone to determine its overall age before consolidating the findings with the radiographic reference atlas to achieve the best match. This review provides strong motivation for automation that automatically identifies and analyzes bone structural changes in X-rays, thereby providing digital evidence, as well as recent developments in vision-based approaches for pediatric endocrinology, growth disorder diagnosis, and forensic medicine applications. The study highlights the detailed characteristics of the datasets and their challenges, and addresses them with appropriate preprocessing strategies. The detailed evaluation metrics for segmentation, detection, classification, and prediction, along with the corresponding mathematical expressions. Exploring various segmentation methods using traditional and DL techniques to capture growth patterns. Regression and classification tasks of ML and DL methodologies are reviewed based on extracted features, and the performance of each model is tabulated and visualized with clusters of similar approaches. Finally, the review concludes by highlighting current challenges and offering recommendations for future development of the ABAE system. This study, with its comprehensive coverage of the field, serves as a fundamental resource for researchers, offering valuable insights to advance precise investigation of skeletal maturity and to narrow the gap between research and application in healthcare.

4 December 2025 in Algorithms
Verified
Journal Article
Open access
PDF file
PMPriyanka M
Priyanka M
MNMANOHAR N
MANOHAR N
SRSuresha R
Suresha R
DRDayananda Rudrappa
Dayananda Rudrappa
Estimating bone age is essential for accurate diagnoses, appropriate care based on biological age, and fairness in legal matters. In the Indian medicolegal context, determining age through a clinical approach involves analyzing multiple joints; however, the traditional method can be tedious and subjective, relying heavily on human expertise, which may lead to biased decisions in age-related legal disputes. Moreover, commonly used radiographs often exhibit pixel-level variations due to heterogeneous contrast, which complicate segmentation tasks and lead to inconsistencies and reduced model performance. The study presents a multifocal region-based symbolic segmentation technique to automatically retain the soft-tissue region that harbors a growth pattern of an ossification center. Experimental results demonstrate an 84.5% Jaccard similarity, an 81.4% Dice coefficient, an 88.3% precision, a 90.0% recall, and a 91.5% pixel accuracy on a novel multifocal dataset of Indian inhabitants. The proposed segmentation technique outperforms U-Net, Attention U-Net, TransU-Net, DeepLabV3+, Adaptive Otsu, and Watershed segmentation in terms of accuracy, indicating strong generalizability across joints and improving reliability. Compared with 86.4% without segmentation, the proposed integration of segmentation with VGG16 classification increases the overall accuracy to 93.8%, demonstrating that target-focused-region processing reduces unnecessary computations and improves feature discrimination without sacrificing accuracy.