Debi Prosad Dogra
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Debi Prosad Dogra has a total of 124 co-authors. Here are the top co-authors by number of shared publications:
- Partha Pratim Roy
Department of Computer Science and Engineering, IIT Roorkee, Roorkee, India
42 shared publications - Arif Ahmed Sekh
UiT The Arctic University of Norway
19 shared publications - Heeseung Choi
Center for Imaging Media Research, Korea Institute of Science and Technology, Seoul 02792, Korea
9 shared publications - Santhosh Kelathodi Kumaran
School of Electrical Sciences, Indian Institute of Technology Bhubaneswar, Odisha, India
9 shared publications - Ig-Jae Kim
Korea Institute of Science and Technology (KIST)
7 shared publications Samarjit Kar
7 shared publications
Latest Publications
- 3 August 2026
- Published by Springer Nature
- p. 672-686
Learning Illumination-Invariant Representations for Vehicle Re-identification
- 3 August 2026
- Published by Springer Nature
- p. 672-686
Learning Illumination-Invariant Representations for Vehicle Re-identification
- 17 July 2026
- Published by Elsevier in Information Sciences
- Vol. 757
Physics-constrained dynamic graph and multi-resolution temporal modeling for traffic forecasting
- 17 July 2026
- Published by Elsevier in Information Sciences
- Vol. 757
Physics-constrained dynamic graph and multi-resolution temporal modeling for traffic forecasting
- Open Access
- 19 May 2026
- Published by Elsevier in SSRN Electronic Journal
Identity-Supervised Graph Transformer with Uncertainty-Calibrated Edges for Vehicle Re-Identification
Vehicle re-identification (Re-ID) aims to match images of the same vehicleacross non-overlapping camera views, a task central to multi-camera visualsurveillance and large-scale video analytics.The core difficulty stems from near-identical inter-instance appearancesand substantial intra-class variation induced by viewpoint and illuminationchanges.Current methods face three fundamental limitations: geometric reasoning relies on manually annotated keypoints, relational graph construction employs deterministic similarity metrics that propagate errors under appearance ambiguity, and inference cost is treated as a fixed quantity regardless of input complexity.In this paper, we propose Keypoint-Aware Graph Transformer Network (KAGT-Net), an identity-supervised graph transformer with uncertainty-calibrated edges, organized as a four-stage hierarchical framework for vehicle Re-ID. We discover geometric structure through an identity-supervised keypoint prediction module that leverages Procrustes consistency across identity-matched pairs as the sole geometric supervision signal,eliminating the need for manual part annotations.We construct a sparse relational graph using uncertainty-calibrated edge probabilities derived from heteroscedastic estimation, suppressing spurious associations under appearance ambiguity.We align cross-camera feature distributions via differentiable optimal transport, coupled with a temporal memory module for identity-consistent embedding aggregation.An adaptive budget controller dynamically adjusts graph sparsity to balanceretrieval accuracy against latency constraints.Experiments on VeRi-776, VehicleID, and VERI-Wild demonstrate consistentimprovements over recent state-of-the-art methods, confirming theeffectiveness of uncertainty-calibrated relational reasoning for fine-grainedvehicle retrieval.
- 19 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessIdentity-Supervised Graph Transformer with Uncertainty-Calibrated Edges for Vehicle Re-Identification
Vehicle re-identification (Re-ID) aims to match images of the same vehicleacross non-overlapping camera views, a task central to multi-camera visualsurveillance and large-scale video analytics.The core difficulty stems from near-identical inter-instance appearancesand substantial intra-class variation induced by viewpoint and illuminationchanges.Current methods face three fundamental limitations: geometric reasoning relies on manually annotated keypoints, relational graph construction employs deterministic similarity metrics that propagate errors under appearance ambiguity, and inference cost is treated as a fixed quantity regardless of input complexity.In this paper, we propose Keypoint-Aware Graph Transformer Network (KAGT-Net), an identity-supervised graph transformer with uncertainty-calibrated edges, organized as a four-stage hierarchical framework for vehicle Re-ID. We discover geometric structure through an identity-supervised keypoint prediction module that leverages Procrustes consistency across identity-matched pairs as the sole geometric supervision signal,eliminating the need for manual part annotations.We construct a sparse relational graph using uncertainty-calibrated edge probabilities derived from heteroscedastic estimation, suppressing spurious associations under appearance ambiguity.We align cross-camera feature distributions via differentiable optimal transport, coupled with a temporal memory module for identity-consistent embedding aggregation.An adaptive budget controller dynamically adjusts graph sparsity to balanceretrieval accuracy against latency constraints.Experiments on VeRi-776, VehicleID, and VERI-Wild demonstrate consistentimprovements over recent state-of-the-art methods, confirming theeffectiveness of uncertainty-calibrated relational reasoning for fine-grainedvehicle retrieval.
- Open Access
- 2 May 2026
- Published by Elsevier in SSRN Electronic Journal
Physics-Constrained Dynamic Graph and Multi-Resolution Temporal Modeling for TrafficForecasting
Spatio-temporal traffic forecasting in Intelligent TransportationSystems (ITS) remains challenging due to four persistentinformation-processing limitations: static graph dependenciesthat obscure evolving network topology, insufficient uncertaintyquantification under heterogeneous sensor reliability, uniform temporal resolution that prevents multi-scale information fusion,and physical constraints enforced in isolation without aunified probabilistic learning formulation.In this paper, we propose \textbf{PRISM-ST} (Physics-informedResilient Intelligence for Multi-scale Spatio-TemporalTransportation), a modular framework integrating domain knowledgeencoding with information-theoretic reliability modeling foradaptive prediction from road sensor measurements.PRISM-ST comprises four jointly optimized modules: an Uncertainty-Aware Spectral-Spatial Block (UASSB) forsensor-reliability-weighted spatial embedding via dual-pathwaygraph convolution; a Unified Dynamics and Structure Learning (UDSL) module enforcingtraffic-theoretic constraints through physics-informed neural ODEsand variational structure learning; a Multi-Resolution Temporal Decomposition (MRTD) module enablingvolatility-guided multi-scale information fusion across temporalscales; and an Adaptive Attention-Memory Routing (AAMR) module that provides the horizon-specific prediction head with mixture-of-experts routing. Extensive evaluation on six benchmark datasets ---PeMS03, PeMS04, PeMS07, PeMS08, GBA, and GLA ---demonstrates an average 5.5\% Mean Absolute Percentage Error (MAPE) reduction and 2.1\% Mean Absolute Error (MAE) reduction in traffic flow prediction over thestrongest baseline per dataset, spanning competitivegraph-based, Transformer-based, physics-informed, anduncertainty-aware baselines.
- 2 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessPhysics-Constrained Dynamic Graph and Multi-Resolution Temporal Modeling for TrafficForecasting
Spatio-temporal traffic forecasting in Intelligent TransportationSystems (ITS) remains challenging due to four persistentinformation-processing limitations: static graph dependenciesthat obscure evolving network topology, insufficient uncertaintyquantification under heterogeneous sensor reliability, uniform temporal resolution that prevents multi-scale information fusion,and physical constraints enforced in isolation without aunified probabilistic learning formulation.In this paper, we propose \textbf{PRISM-ST} (Physics-informedResilient Intelligence for Multi-scale Spatio-TemporalTransportation), a modular framework integrating domain knowledgeencoding with information-theoretic reliability modeling foradaptive prediction from road sensor measurements.PRISM-ST comprises four jointly optimized modules: an Uncertainty-Aware Spectral-Spatial Block (UASSB) forsensor-reliability-weighted spatial embedding via dual-pathwaygraph convolution; a Unified Dynamics and Structure Learning (UDSL) module enforcingtraffic-theoretic constraints through physics-informed neural ODEsand variational structure learning; a Multi-Resolution Temporal Decomposition (MRTD) module enablingvolatility-guided multi-scale information fusion across temporalscales; and an Adaptive Attention-Memory Routing (AAMR) module that provides the horizon-specific prediction head with mixture-of-experts routing. Extensive evaluation on six benchmark datasets ---PeMS03, PeMS04, PeMS07, PeMS08, GBA, and GLA ---demonstrates an average 5.5\% Mean Absolute Percentage Error (MAPE) reduction and 2.1\% Mean Absolute Error (MAE) reduction in traffic flow prediction over thestrongest baseline per dataset, spanning competitivegraph-based, Transformer-based, physics-informed, anduncertainty-aware baselines.
- Open Access
- 14 February 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Machine learning-based heat flux estimation from high-speed video during saturated pool boiling over vertical tube
Saturated pool boiling over vertical tube has significant importance in various practical applications specifically in the safety of nuclear power plant. The process of boiling is characterized by the formation of bubbles within the liquid. These bubbles are composed of vapour of the substance. The formation and growth of bubbles are influenced by various factors which directly regulate the nucleation sites in the liquid. The formation of nucleation sites is directly linked to the heat flux and heat transfer efficiency of the device. The accurate prediction of these parameter is quite important for the equipment design, safety, reliability. Here, we used machine learning approach that correlates high-quality high-speed imaging on dynamic bubbles at different heat flux. The framework leverages cutting-edge deep learning models including convolutional neural networks and object detection algorithms to automatically extract both hierarchical and physics-based features. By training on these features, our model learns physical boiling laws that statistically describe the manner in which bubbles nucleate, coalesce, and depart under boiling conditions, enabling in situ boiling heat flux prediction. The proposed method achieves an average heat flux prediction error of approximately 6%, while maintaining an overall classification accuracy of 88% across all heat flux levels. Our framework offers an automated, learning-based, alternative to conventional boiling heat transfer metrology.
- 14 February 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Open AccessMachine learning-based heat flux estimation from high-speed video during saturated pool boiling over vertical tube
Saturated pool boiling over vertical tube has significant importance in various practical applications specifically in the safety of nuclear power plant. The process of boiling is characterized by the formation of bubbles within the liquid. These bubbles are composed of vapour of the substance. The formation and growth of bubbles are influenced by various factors which directly regulate the nucleation sites in the liquid. The formation of nucleation sites is directly linked to the heat flux and heat transfer efficiency of the device. The accurate prediction of these parameter is quite important for the equipment design, safety, reliability. Here, we used machine learning approach that correlates high-quality high-speed imaging on dynamic bubbles at different heat flux. The framework leverages cutting-edge deep learning models including convolutional neural networks and object detection algorithms to automatically extract both hierarchical and physics-based features. By training on these features, our model learns physical boiling laws that statistically describe the manner in which bubbles nucleate, coalesce, and depart under boiling conditions, enabling in situ boiling heat flux prediction. The proposed method achieves an average heat flux prediction error of approximately 6%, while maintaining an overall classification accuracy of 88% across all heat flux levels. Our framework offers an automated, learning-based, alternative to conventional boiling heat transfer metrology.
- Open Access
- 23 December 2025
- Published by Elsevier in Data in Brief
- Vol. 64
Pixel-level annotations of context region for trustworthy diagnosis of Breast Lymph Node Metastasis from Histopathological WSI
- 23 December 2025
- Published by Elsevier in Data in Brief
- Vol. 64
Open AccessPixel-level annotations of context region for trustworthy diagnosis of Breast Lymph Node Metastasis from Histopathological WSI
- Open Access
- 27 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 3182-3191
Can Person-Level Attributes Improve Group Re-Identification?
- 27 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 3182-3191
Open AccessCan Person-Level Attributes Improve Group Re-Identification?
- 24 July 2025
- Published by Springer Nature
- p. 423-438
Prediction of Short-Term Traffic Flow Using a Hybrid Architecture
- 24 July 2025
- Published by Springer Nature
- p. 423-438
Prediction of Short-Term Traffic Flow Using a Hybrid Architecture
- 1 July 2025
- Published by Elsevier in Biomedical Signal Processing and Control
- Vol. 110
Feature reweighting for EEG-based motor imagery classification
- 1 July 2025
- Published by Elsevier in Biomedical Signal Processing and Control
- Vol. 110
Feature reweighting for EEG-based motor imagery classification
- 25 April 2025
- Published by Springer Nature
- p. 335-349
Prediction of Urban Traffic Patterns by Analyzing Real-World Videos Using Clustering and Markov Model
- 25 April 2025
- Published by Springer Nature
- p. 335-349
Prediction of Urban Traffic Patterns by Analyzing Real-World Videos Using Clustering and Markov Model