Sabari Nathan

Sabari Nathan

Computer Vision Research Engineer

Deep Learning & Computer Vision Specialist

Email Google Scholar GitHub LinkedIn CV

Computer Vision Research Engineer with 10+ years of experience in deep learning, cultural heritage preservation, and practical AI applications. Published 25+ papers in top-tier conferences and journals including CVPR, ECCV, Heritage Science, and Journal of Food Engineering. Proven competition champion with 1st place victory in SSBC 2025 and multiple top-3 finishes in prestigious computer vision challenges including CVPR competitions (thermal super-resolution, skeleton segmentation), Facebook OpenEDS, and Kaggle Gold Medal (Top 1%). Expertise spans face anti-spoofing systems, heritage image classification, food technology AI, UAV-based infrastructure monitoring, and mobile camera enhancement. Strong focus on creating real-world impact through interdisciplinary research spanning computer vision, cultural studies, and industrial applications.

Research

My research spans multiple domains of computer vision and AI applications, with a strong focus on cultural heritage preservation, practical AI systems, and real-world problem solving. I specialize in developing deep learning solutions for heritage image classification (Kolkata monuments, Tamil Kolam art), face anti-spoofing and security systems, food technology and agricultural applications (freshness prediction, medicinal plant identification), transportation infrastructure monitoring (UAV-based railroad analysis), and mobile camera enhancement technologies. My work emphasizes bridging academic research with industrial applications and cultural preservation.

How can computer vision technologies solve diverse real-world challenges while preserving cultural heritage and advancing practical AI applications across multiple industries?

Publications

[Face Anti-Spoofing]

Multiattention-Net: A Novel Approach to Face Anti-Spoofing with Modified Squeezed Residual Blocks

Sabari Nathan, M Parisa Beham, A Nagaraj, S Roomi

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024

We present Multiattention-Net, a novel deep learning architecture for face anti-spoofing that combines attention mechanisms with modified squeezed residual blocks. Our approach achieves state-of-the-art performance in detecting presentation attacks while maintaining computational efficiency for real-time applications.

Paper Code Demo
[Heritage Classification]

MonuNet: A High Performance Deep Learning Network for Kolkata Heritage Image Classification

A Sasithradevi, B Chanthini, T Subbulakshmi, P Prakash

Heritage Science, Springer, Vol. 12, pp. 1-14, 2024

MonuNet introduces a specialized deep learning architecture for classifying Kolkata heritage monuments and architectural elements. Our network achieves superior performance in heritage image recognition, contributing to digital preservation of Indian cultural assets and supporting heritage tourism and conservation efforts.

Paper Dataset Code
[Food Freshness AI]

Is Human Perception Reliable? Toward Illumination Robust Food Freshness Prediction from Food Appearance

D Wang, S Sethu, S Nathan, Z Li, VJ Hogan, C Ni, S Zhang, HS Seo

Journal of Food Engineering, Vol. 381, 112179, 2024

We develop an illumination-robust AI system for predicting food freshness from visual appearance, using lettuce as a case study. Our approach outperforms human perception in consistency and accuracy, offering practical applications for food quality assessment in retail and agricultural industries.

Paper Dataset Code
[Tamil Heritage Art]

KolamNetV2: Efficient Attention-Based Deep Learning Network for Tamil Heritage Art-Kolam Classification

A Sasithradevi, Sabarinathan, S Shoba, SMM Roomi, P Prakash

Heritage Science, Springer, Vol. 12, pp. 60, 2024

KolamNetV2 presents an enhanced attention-based architecture for classifying traditional Tamil Kolam art patterns. This work advances digital preservation of South Indian cultural heritage through improved accuracy and efficiency in recognizing complex geometric patterns and artistic traditions.

Paper Dataset Code
[Railroad UAV System]

Lightweight Railroad Semantic Segmentation Network and Distance Estimation for Railroad UAV Images

RS Rampriya, S Nathan, R Suganya, SB Prathiba, PS Perumal, W Wang

Engineering Applications of Artificial Intelligence, Vol. 134, 108620, 2023

We develop a lightweight neural network for real-time railroad semantic segmentation and distance estimation from UAV imagery. Our system enables automated railway infrastructure monitoring and maintenance planning, contributing to safer and more efficient railway operations through computer vision.

Paper Code Dataset
[Mobile Camera Enhancement]

Real-Time Under-Display Cameras Image Restoration and HDR on Mobile Devices

MV Conde, F Vasluianu, S Nathan, R Timofte

European Conference on Computer Vision (ECCV), pp. 747-762, 2022

We present a real-time solution for under-display camera image restoration and HDR processing optimized for mobile devices. Our approach addresses the unique challenges of under-display camera technology, enabling high-quality imaging through efficient deep learning methods suitable for smartphone deployment.

Paper Code Demo

News

Jan 2025
🥇 WON 1st Place in Synthetic Dataset Track at 9th Sclera Segmentation and Benchmarking Competition (SSBC 2025)!
Jun 2024
🏆 Secured 7th place in Demosaic for Hybrid-EVS Camera competition at MIPI@CVPR 2024!
Nov 2024
Face anti-spoofing paper accepted at CVPR 2024! Multiattention-Net achieves state-of-the-art performance
Oct 2024
Two heritage preservation papers published in Heritage Science: MonuNet for Kolkata monuments and KolamNetV2
Sep 2024
Food freshness prediction research published in Journal of Food Engineering - breakthrough in AI-assisted food quality assessment

Education

[Deakin University]

Master of Data Science (Global)

Deakin University, Australia

Graduate Degree • Data Science Specialization

Comprehensive graduate program focusing on advanced data science methodologies, machine learning algorithms, statistical analysis, and big data technologies. Global program format providing international perspective on data science applications across various industries and research domains.

[UT McCombs]

Postgraduate Program in Artificial Intelligence & Machine Learning

Texas McCombs School of Business, University of Texas at Austin

Postgraduate Certificate • GPA: 3.82/4.0

Advanced postgraduate program at one of the top business schools focusing on AI and ML applications in business contexts. Curriculum covered deep learning, computer vision, natural language processing, and strategic implementation of AI technologies in enterprise environments.

[TCE Engineering]

Bachelor of Engineering (B.E.) - Electronics and Communications Engineering

Thiagarajar College of Engineering, India

Undergraduate Degree • CGPA: 8.78/10.0

Strong foundational engineering degree with focus on electronics, communications systems, signal processing, and digital systems. Excellent academic performance providing solid technical foundation for advanced computer vision and machine learning research.

Selected Additional Publications

Medical & Agricultural AI

  • SIMPD Net: South Indian Medicinal Plants Dataset (IEEE SILCON 2024)
  • Polyp Segmentation: Multi-supervision Net with Attention (MAI 2022)
  • Pesticide Residue Estimation: 3D Squeeze Excitation-Residual Network (ICCVG 2022)

Heritage & Cultural Preservation

  • KolamNet: Attention-based Tamil Kolam Classification (ICCVG 2022)
  • MonuNet: Kolkata Heritage Image Classification (Heritage Science 2024)
  • KolamNetV2: Enhanced Tamil Heritage Art Classification (Heritage Science 2024)

Computer Vision & Mobile Applications

  • Depth-guided Relighting: Lightweight Deep Learning Method (Journal of Imaging 2023)
  • Instagram Filter Removal: Recurrent Residual Network (ICCVG 2022)
  • Under-display Cameras: Real-time Image Restoration (ECCV 2022)

Transportation & Infrastructure

  • Railroad Obstacle Detection: Deep Neural Networks (Applied AI 2022)
  • Railroad Segmentation: Modified UNet Architecture (ICCIDE 2021)
  • UAV Railroad Analysis: Lightweight Segmentation Network (Eng. Apps. AI 2023)

Academic Service

Research Specializations

  • Biometric Systems & Sclera Segmentation
  • Cultural Heritage & Art Classification
  • Face Anti-Spoofing & Biometric Security
  • Food Technology & Agricultural AI
  • UAV-based Infrastructure Monitoring
  • Mobile Camera Enhancement Technologies
  • Medical Image Analysis & Segmentation

Reviewer

  • CVPR, ICCV, ECCV
  • NeurIPS, ICML, ICLR
  • TPAMI, IJCV
  • Nature Machine Intelligence
  • IEEE TIP, IEEE TMM

Workshops

  • CVPR Workshop on Computational Photography
  • ICCV Workshop on Mobile AI
  • NeurIPS Workshop on ML for Systems
  • ECCV Workshop on Advances in Image Manipulation

Competition Awards & Recognition

Contact

Get in Touch

Email: your.email@domain.com

Office: Room 123, Computer Science Building

University/Institution Name

City, Country