Ayan Banerjee

Kunal Purkayastha, Ayan Banerjee, Josep Llados et al.

In Document Understanding, the challenge of reconstructing damaged, occluded, or incomplete text remains a critical yet unexplored problem. Subsequent document understanding tasks can benefit from a document reconstruction process. In response, this paper presents a novel unified pipeline combining state-of-the-art Optical Character Recognition (OCR), advanced image analysis, masked language modeling, and diffusion-based models to restore and reconstruct text while preserving visual integrity. We create a synthetic dataset of 30{,}078 degraded document images that simulates diverse document degradation scenarios, setting a benchmark for restoration tasks. Our pipeline detects and recognizes text, identifies degradation with an occlusion detector, and uses an inpainting model for semantically coherent reconstruction. A diffusion-based module seamlessly reintegrates text, matching font, size, and alignment. To evaluate restoration quality, we propose a Unified Context Similarity Metric (UCSM), incorporating edit, semantic, and length similarities with a contextual predictability measure that penalizes deviations when the correct text is contextually obvious. Our work advances document restoration, benefiting archival research and digital preservation while setting a new standard for text reconstruction. The OPRB dataset and code are available at \href{https://huggingface.co/datasets/kpurkayastha/OPRB}{Hugging Face} and \href{https://github.com/kunalpurkayastha/DocRevive}{Github} respectively.

Ayan Banerjee, Payal Kamboj, Aranyak Maity et al.

Non-linearities in simulation arise from the time variance in wireless mobile networks when integrated with human in the loop, human in the plant (HIL-HIP) physical systems under dynamic contexts, leading to simulation slowdown. Time variance is handled by deriving a series of piece wise linear time invariant simulations (PLIS) in intervals, which are then concatenated in time domain. In this paper, we conduct a formal analysis of the impact of discretizing time-varying components in wireless network-controlled HIL-HIP systems on simulation accuracy and speedup, and evaluate trade-offs with reliable guarantees. We develop an accurate simulation framework for an artificial pancreas wireless network system that controls blood glucose in Type 1 Diabetes patients with time varying properties such as physiological changes associated with psychological stress and meal patterns. PLIS approach achieves accurate simulation with greater than 2.1 times speedup than a non-linear system simulation for the given dataset.

Farhat Shaikh, Ayan Banerjee, Sandeep Gupta

We introduce EMMA, a physics-informed multimodal framework that recovers all identifiable dynamical parameters of a system directly from raw video, audio, and image-based time-series observations. Unlike prior video-only approaches that struggle with occluded states, hidden actuation inputs, or assumptions about known initial conditions and coordinate frames, EMMA performs joint inference of explicit parameters, implicit dynamical components, and calibration invariants within a unified continuous-time model. EMMA leverages a Liquid Time-Constant (LTC) network to learn latent dynamics from heterogeneous modalities while a physics-constrained loss enforces consistency with the governing differential equations. A unified feature pipeline enables consistent alignment across video trajectories, acoustic signatures, and chart-derived measurements, allowing EMMA to estimate parameters under forced, implicit, and multivariate dynamics without requiring segmentation masks, differentiable rendering, or specialized sensors. Across 100+ scenarios including five standard dynamical benchmarks (75 Delfys videos), real-world rover and quadrotor systems with hidden inputs, and simulation-chart case studies spanning biological and chaotic systems, EMMA delivers robust multi-parameter recovery and significantly outperforms existing single-modality and equation-discovery baselines. Our results establish EMMA as a general, scalable solution for physics-consistent model extraction from opportunistic multimodal data. Code and data are available at: https://github.com/ImpactLabASU/EMMA-CVPR2026

Aranyak Maity, Ayan Banerjee, Sandeep Gupta

Unknown-unknowns are operational scenarios in systems that are not accounted for in the design and test phase. In such scenarios, the operational behavior of the Human-in-loop (HIL) Human-in-Plant (HIP) systems is not guaranteed to meet requirements such as safety and efficacy. We propose a novel framework for analyzing the operational output characteristics of safety-critical HIL-HIP systems that can discover unknown-unknown scenarios and evaluate potential safety hazards. We propose dynamics-induced hybrid recurrent neural networks (DiH-RNN) to mine a physics-guided surrogate model (PGSM) that checks for deviation of the cyber-physical system (CPS) from safety-certified operational characteristics. The PGSM enables early detection of unknown-unknowns based on the physical laws governing the system. We demonstrate the detection of operational changes in an Artificial Pancreas(AP) due to unknown insulin cartridge errors.

Sameena Hossain, Payal Kamboj, Aranyak Maity et al.

Gestures that share similarities in their forms and are related in their meanings, should be easier for learners to recognize and incorporate into their existing lexicon. In that regard, to be more readily accepted as standard by the Deaf and Hard of Hearing community, technical gestures in American Sign Language (ASL) will optimally share similar in forms with their lexical neighbors. We utilize a lexical database of ASL, ASL-LEX, to identify lexical relations within a set of technical gestures. We use automated identification for 3 unique sub-lexical properties in ASL- location, handshape and movement. EdGCon assigned an iconicity rating based on the lexical property similarities of the new gesture with an existing set of technical gestures and the relatedness of the meaning of the new technical word to that of the existing set of technical words. We collected 30 ad hoc crowdsourced technical gestures from different internet websites and tested them against 31 gestures from the DeafTEC technical corpus. We found that EdGCon was able to correctly auto-assign the iconicity ratings 80.76% of the time.

Payal Kamboj, Ayan Banerjee, Sandeep K. S. Gupta et al.

Surgical disconnection of Seizure Onset Zones (SOZs) at an early age is an effective treatment for Pharmaco-Resistant Epilepsy (PRE). Pre-surgical localization of SOZs with intra-cranial EEG (iEEG) requires safe and effective depth electrode placement. Resting-state functional Magnetic Resonance Imaging (rs-fMRI) combined with signal decoupling using independent component (IC) analysis has shown promising SOZ localization capability that guides iEEG lead placement. However, SOZ ICs identification requires manual expert sorting of 100s of ICs per patient by the surgical team which limits the reproducibility and availability of this pre-surgical screening. Automated approaches for SOZ IC identification using rs-fMRI may use deep learning (DL) that encodes intricacies of brain networks from scarcely available pediatric data but has low precision, or shallow learning (SL) expert rule-based inference approaches that are incapable of encoding the full spectrum of spatial features. This paper proposes DeepXSOZ that exploits the synergy between DL based spatial feature and SL based expert knowledge encoding to overcome performance drawbacks of these strategies applied in isolation. DeepXSOZ is an expert-in-the-loop IC sorting technique that a) can be configured to either significantly reduce expert sorting workload or operate with high sensitivity based on expertise of the surgical team and b) can potentially enable the usage of rs-fMRI as a low cost outpatient pre-surgical screening tool. Comparison with state-of-art on 52 children with PRE shows that DeepXSOZ achieves sensitivity of 89.79%, precision of 93.6% and accuracy of 84.6%, and reduces sorting effort by 6.7-fold. Knowledge level ablation studies show a pathway towards maximizing patient outcomes while optimizing the machine-expert collaboration for various scenarios.

Payal Kamboj, Ayan Banerjee, Sandeep K. S. Gupta

Knowledge transfer across sensing technology is a novel concept that has been recently explored in many application domains, including gesture-based human computer interaction. The main aim is to gather semantic or data driven information from a source technology to classify / recognize instances of unseen classes in the target technology. The primary challenge is the significant difference in dimensionality and distribution of feature sets between the source and the target technologies. In this paper, we propose TRANSFER, a generic framework for knowledge transfer between a source and a target technology. TRANSFER uses a language-based representation of a hand gesture, which captures a temporal combination of concepts such as handshape, location, and movement that are semantically related to the meaning of a word. By utilizing a pre-specified syntactic structure and tokenizer, TRANSFER segments a hand gesture into tokens and identifies individual components using a token recognizer. The tokenizer in this language-based recognition system abstracts the low-level technology-specific characteristics to the machine interface, enabling the design of a discriminator that learns technology-invariant features essential for recognition of gestures in both source and target technologies. We demonstrate the usage of TRANSFER for three different scenarios: a) transferring knowledge across technology by learning gesture models from video and recognizing gestures using WiFi, b) transferring knowledge from video to accelerometer, and d) transferring knowledge from accelerometer to WiFi signals.

Alloy Das, Sanket Biswas, Ayan Banerjee et al.

The adaptation capability to a wide range of domains is crucial for scene text spotting models when deployed to real-world conditions. However, existing state-of-the-art (SOTA) approaches usually incorporate scene text detection and recognition simply by pretraining on natural scene text datasets, which do not directly exploit the intermediate feature representations between multiple domains. Here, we investigate the problem of domain-adaptive scene text spotting, i.e., training a model on multi-domain source data such that it can directly adapt to target domains rather than being specialized for a specific domain or scenario. Further, we investigate a transformer baseline called Swin-TESTR to focus on solving scene-text spotting for both regular and arbitrary-shaped scene text along with an exhaustive evaluation. The results clearly demonstrate the potential of intermediate representations to achieve significant performance on text spotting benchmarks across multiple domains (e.g. language, synth-to-real, and documents). both in terms of accuracy and efficiency.

Bin Xu, Ayan Banerjee, Sandeep K. S. Gupta

Model Recovery (MR) enables safe, explainable decision making in mission-critical autonomous systems (MCAS) by learning governing dynamical equations, but its deployment on edge devices is hindered by the iterative nature of neural ordinary differential equations (NODEs), which are inefficient on FPGAs. Memory and energy consumption are the main concerns when applying MR on edge devices for real-time operation. We propose MERINDA, a novel FPGA-accelerated MR framework that replaces iterative solvers with a parallelizable neural architecture equivalent to NODEs. MERINDA achieves nearly 11x lower DRAM usage and 2.2x faster runtime compared to mobile GPUs. Experiments reveal an inverse relationship between memory and energy at fixed accuracy, highlighting MERINDA's suitability for resource-constrained, real-time MCAS.

Bernard Ngabonziza, Ayan Banerjee, Sandeep K. S. Gupta

In recent years, Human-centric cyber-physical systems have increasingly involved artificial intelligence to enable knowledge extraction from sensor-collected data. Examples include medical monitoring and control systems, as well as autonomous cars. Such systems are intended to operate according to the protocols and guidelines for regular system operations. However, in many scenarios, such as closed-loop blood glucose control for Type 1 diabetics, self-driving cars, and monitoring systems for stroke diagnosis. The operations of such AI-enabled human-centric applications can expose them to cases for which their operational mode may be uncertain, for instance, resulting from the interactions with a human with the system. Such cases, in which the system is in uncertain conditions, can violate the system's safety and security requirements. This paper will discuss operational deviations that can lead these systems to operate in unknown conditions. We will then create a framework to evaluate different strategies for ensuring the safety and security of AI-enabled human-centric cyber-physical systems in operation deployment. Then, as an example, we show a personalized image-based novel technique for detecting the non-announcement of meals in closed-loop blood glucose control for Type 1 diabetics.

Bernard Ngabonziza, Ayan Banerjee, Sandeep K. S. Gupta

Human centric critical systems are increasingly involving artificial intelligence to enable knowledge extraction from sensor collected data. Examples include medical monitoring and control systems, gesture based human computer interaction systems, and autonomous cars. Such systems are intended to operate for a long term potentially for a lifetime in many scenarios such as closed loop blood glucose control for Type 1 diabetics, self-driving cars, and monitoting systems for stroke diagnosis, and rehabilitation. Long term operation of such AI enabled human centric applications can expose them to corner cases for which their operation is may be uncertain. This can be due to many reasons such as inherent flaws in the design, limited resources for testing, inherent computational limitations of the testing methodology, or unknown use cases resulting from human interaction with the system. Such untested corner cases or cases for which the system performance is uncertain can lead to violations in the safety, sustainability, and security requirements of the system. In this paper, we analyze the existing techniques for safety, sustainability, and security analysis of an AI enabled human centric control system and discuss their limitations for testing the system for long term use in practice. We then propose personalized model based solutions for potentially eliminating such limitations.

Ayan Banerjee, Komandoor Srivathsan, Sandeep K. S. Gupta

A desirable property of any deployed artificial intelligence is generalization across domains, i.e. data generation distribution under a specific acquisition condition. In medical imagining applications the most coveted property for effective deployment is Single Domain Generalization (SDG), which addresses the challenge of training a model on a single domain to ensure it generalizes well to unseen target domains. In multi-center studies, differences in scanners and imaging protocols introduce domain shifts that exacerbate variability in rare class characteristics. This paper presents our experience on SDG in real life deployment for two exemplary medical imaging case studies on seizure onset zone detection using fMRI data, and stress electrocardiogram based coronary artery detection. Utilizing the commonly used application of diabetic retinopathy, we first demonstrate that state-of-the-art SDG techniques fail to achieve generalized performance across data domains. We then develop a generic expert knowledge integrated deep learning technique DL+EKE and instantiate it for the DR application and show that DL+EKE outperforms SOTA SDG methods on DR. We then deploy instances of DL+EKE technique on the two real world examples of stress ECG and resting state (rs)-fMRI and discuss issues faced with SDG techniques.

Ayan Banerjee, Nityanand Mathur, Josep Lladós et al.

Generating VectorArt from text prompts is a challenging vision task, requiring diverse yet realistic depictions of the seen as well as unseen entities. However, existing research has been mostly limited to the generation of single objects, rather than comprehensive scenes comprising multiple elements. In response, this work introduces SVGCraft, a novel end-to-end framework for the creation of vector graphics depicting entire scenes from textual descriptions. Utilizing a pre-trained LLM for layout generation from text prompts, this framework introduces a technique for producing masked latents in specified bounding boxes for accurate object placement. It introduces a fusion mechanism for integrating attention maps and employs a diffusion U-Net for coherent composition, speeding up the drawing process. The resulting SVG is optimized using a pre-trained encoder and LPIPS loss with opacity modulation to maximize similarity. Additionally, this work explores the potential of primitive shapes in facilitating canvas completion in constrained environments. Through both qualitative and quantitative assessments, SVGCraft is demonstrated to surpass prior works in abstraction, recognizability, and detail, as evidenced by its performance metrics (CLIP-T: 0.4563, Cosine Similarity: 0.6342, Confusion: 0.66, Aesthetic: 6.7832). The code will be available at https://github.com/ayanban011/SVGCraft.

Ayan Banerjee, Sanket Biswas, Josep Lladós et al.

Object detection in documents is a key step to automate the structural elements identification process in a digital or scanned document through understanding the hierarchical structure and relationships between different elements. Large and complex models, while achieving high accuracy, can be computationally expensive and memory-intensive, making them impractical for deployment on resource constrained devices. Knowledge distillation allows us to create small and more efficient models that retain much of the performance of their larger counterparts. Here we present a graph-based knowledge distillation framework to correctly identify and localize the document objects in a document image. Here, we design a structured graph with nodes containing proposal-level features and edges representing the relationship between the different proposal regions. Also, to reduce text bias an adaptive node sampling strategy is designed to prune the weight distribution and put more weightage on non-text nodes. We encode the complete graph as a knowledge representation and transfer it from the teacher to the student through the proposed distillation loss by effectively capturing both local and global information concurrently. Extensive experimentation on competitive benchmarks demonstrates that the proposed framework outperforms the current state-of-the-art approaches. The code will be available at: https://github.com/ayanban011/GraphKD.

Ayan Banerjee, Sanket Biswas, Josep Lladós et al.

Instance-level segmentation of documents consists in assigning a class-aware and instance-aware label to each pixel of the image. It is a key step in document parsing for their understanding. In this paper, we present a unified transformer encoder-decoder architecture for en-to-end instance segmentation of complex layouts in document images. The method adapts a contrastive training with a mixed query selection for anchor initialization in the decoder. Later on, it performs a dot product between the obtained query embeddings and the pixel embedding map (coming from the encoder) for semantic reasoning. Extensive experimentation on competitive benchmarks like PubLayNet, PRIMA, Historical Japanese (HJ), and TableBank demonstrate that our model with SwinL backbone achieves better segmentation performance than the existing state-of-the-art approaches with the average precision of \textbf{93.72}, \textbf{54.39}, \textbf{84.65} and \textbf{98.04} respectively under one billion parameters. The code is made publicly available at: \href{https://github.com/ayanban011/SwinDocSegmenter}{github.com/ayanban011/SwinDocSegmenter}

Subhajit Maity, Sanket Biswas, Siladittya Manna et al.

Document layout analysis is a known problem to the documents research community and has been vastly explored yielding a multitude of solutions ranging from text mining, and recognition to graph-based representation, visual feature extraction, etc. However, most of the existing works have ignored the crucial fact regarding the scarcity of labeled data. With growing internet connectivity to personal life, an enormous amount of documents had been available in the public domain and thus making data annotation a tedious task. We address this challenge using self-supervision and unlike, the few existing self-supervised document segmentation approaches which use text mining and textual labels, we use a complete vision-based approach in pre-training without any ground-truth label or its derivative. Instead, we generate pseudo-layouts from the document images to pre-train an image encoder to learn the document object representation and localization in a self-supervised framework before fine-tuning it with an object detection model. We show that our pipeline sets a new benchmark in this context and performs at par with the existing methods and the supervised counterparts, if not outperforms. The code is made publicly available at: https://github.com/MaitySubhajit/SelfDocSeg

Midhat Urooj, Ayan Banerjee, Farhat Shaikh et al.

Accurate and interpretable image-based diagnosis remains a fundamental challenge in medical AI, particularly un- der domain shifts and rare-class conditions. Deep learning mod- els often struggle with real-world distribution changes, exhibit bias against infrequent pathologies, and lack the transparency required for deployment in safety-critical clinical environments. We introduce MedXAI (An Explainable Framework for Med- ical Imaging Classification), a unified expert knowledge based framework that integrates deep vision models with clinician- derived expert knowledge to improve generalization, reduce rare- class bias, and provide human-understandable explanations by localizing the relevant diagnostic features rather than relying on technical post-hoc methods (e.g., Saliency Maps, LIME). We evaluate MedXAI across heterogeneous modalities on two challenging tasks: (i) Seizure Onset Zone localization from resting-state fMRI, and (ii) Diabetic Retinopathy grading. Ex periments on ten multicenter datasets show consistent gains, including a 3% improvement in cross-domain generalization and a 10% improvmnet in F1 score of rare class, substantially outperforming strong deep learning baselines. Ablations confirm that the symbolic components act as effective clinical priors and regularizers, improving robustness under distribution shift. MedXAI delivers clinically aligned explanations while achieving superior in-domain and cross-domain performance, particularly for rare diseases in multimodal medical AI.

Bin Xu, Ayan Banerjee, Sandeep Gupta

Physical AI at the edge -- enabling autonomous systems to understand and predict real-world dynamics in real time -- requires hardware-efficient learning and inference. Model recovery (MR), which identifies governing equations from sensor data, is a key primitive for safe and explainable monitoring in mission-critical autonomous systems operating under strict latency, compute, and power constraints. However, state-of-the-art MR methods (e.g., EMILY and PINN+SR) rely on Neural ODE formulations that require iterative solvers and are difficult to accelerate efficiently on edge hardware. We present \textbf{MERINDA} (Model Recovery in Reconfigurable Dynamic Architecture), an FPGA-accelerated MR framework designed to make physical AI practical on resource-constrained devices. MERINDA replaces expensive Neural ODE components with a hardware-friendly formulation that combines (i) GRU-based discretized dynamics, (ii) dense inverse-ODE layers, (iii) sparsity-driven dropout, and (iv) lightweight ODE solvers. The resulting computation is structured for streaming parallelism, enabling critical kernels to be fully parallelized on the FPGA. Across four benchmark nonlinear dynamical systems, MERINDA delivers substantial gains over GPU implementations: \textbf{114$\times$ lower energy} (434~J vs.\ 49{,}375~J), \textbf{28$\times$ smaller memory footprint} (214~MB vs.\ 6{,}118~MB), and \textbf{1.68$\times$ faster training}, while matching state-of-the-art model-recovery accuracy. These results demonstrate that MERINDA can bring accurate, explainable MR to the edge for real-time monitoring of autonomous systems.

Midhat Urooj, Ayan Banerjee, Sandeep Gupta

Explainability domain generalization and rare class reliability are critical challenges in medical AI where deep models often fail under real world distribution shifts and exhibit bias against infrequent clinical conditions This paper introduces XAIMeD an explainable medical AI framework that integrates clinically accurate expert knowledge into deep learning through a unified neuro symbolic architecture XAIMeD is designed to improve robustness under distribution shift enhance rare class sensitivity and deliver transparent clinically aligned interpretations The framework encodes clinical expertise as logical connectives over atomic medical propositions transforming them into machine checkable class specific rules Their diagnostic utility is quantified through weighted feature satisfaction scores enabling a symbolic reasoning branch that complements neural predictions A confidence weighted fusion integrates symbolic and deep outputs while a Hunt inspired adaptive routing mechanism guided by Entropy Imbalance Gain EIG and Rare Class Gini mitigates class imbalance high intra class variability and uncertainty We evaluate XAIMeD across diverse modalities on four challenging tasks i Seizure Onset Zone SOZ localization from rs fMRI ii Diabetic Retinopathy grading across 6 multicenter datasets demonstrate substantial performance improvements including 6 percent gains in cross domain generalization and a 10 percent improved rare class F1 score far outperforming state of the art deep learning baselines Ablation studies confirm that the clinically grounded symbolic components act as effective regularizers ensuring robustness to distribution shifts XAIMeD thus provides a principled clinically faithful and interpretable approach to multimodal medical AI.

Payal Kamboj, Ayan Banerjee, Varina L. Boerwinkle et al.

We evaluated whether integration of expert guidance on seizure onset zone (SOZ) identification from resting state functional MRI (rs-fMRI) connectomics combined with deep learning (DL) techniques enhances the SOZ delineation in patients with refractory epilepsy (RE), compared to utilizing DL alone. Rs-fMRI were collected from 52 children with RE who had subsequently undergone ic-EEG and then, if indicated, surgery for seizure control (n = 25). The resting state functional connectomics data were previously independently classified by two expert epileptologists, as indicative of measurement noise, typical resting state network connectivity, or SOZ. An expert knowledge integrated deep network was trained on functional connectomics data to identify SOZ. Expert knowledge integrated with DL showed a SOZ localization accuracy of 84.8& and F1 score, harmonic mean of positive predictive value and sensitivity, of 91.7%. Conversely, a DL only model yielded an accuracy of less than 50% (F1 score 63%). Activations that initiate in gray matter, extend through white matter and end in vascular regions are seen as the most discriminative expert identified SOZ characteristics. Integration of expert knowledge of functional connectomics can not only enhance the performance of DL in localizing SOZ in RE, but also lead toward potentially useful explanations of prevalent co-activation patterns in SOZ. RE with surgical outcomes and pre-operative rs-fMRI studies can yield expert knowledge most salient for SOZ identification.

Ayan Banerjee, Aranyak Maity, Payal Kamboj et al.

We explore the usage of large language models (LLM) in human-in-the-loop human-in-the-plant cyber-physical systems (CPS) to translate a high-level prompt into a personalized plan of actions, and subsequently convert that plan into a grounded inference of sequential decision-making automated by a real-world CPS controller to achieve a control goal. We show that it is relatively straightforward to contextualize an LLM so it can generate domain-specific plans. However, these plans may be infeasible for the physical system to execute or the plan may be unsafe for human users. To address this, we propose CPS-LLM, an LLM retrained using an instruction tuning framework, which ensures that generated plans not only align with the physical system dynamics of the CPS but are also safe for human users. The CPS-LLM consists of two innovative components: a) a liquid time constant neural network-based physical dynamics coefficient estimator that can derive coefficients of dynamical models with some unmeasured state variables; b) the model coefficients are then used to train an LLM with prompts embodied with traces from the dynamical system and the corresponding model coefficients. We show that when the CPS-LLM is integrated with a contextualized chatbot such as BARD it can generate feasible and safe plans to manage external events such as meals for automated insulin delivery systems used by Type 1 Diabetes subjects.

Ayan Banerjee, Payal Kamboj, Sandeep Gupta

Precision medicine is a promising approach for accessible disease diagnosis and personalized intervention planning in high-mortality diseases such as coronary artery disease (CAD), drug-resistant epilepsy (DRE), and chronic illnesses like Type 1 diabetes (T1D). By leveraging artificial intelligence (AI), precision medicine tailors diagnosis and treatment solutions to individual patients by explicitly modeling variance in pathophysiology. However, the adoption of AI in medical applications faces significant challenges, including poor generalizability across centers, demographics, and comorbidities, limited explainability in clinical terms, and a lack of trust in ethical decision-making. This paper proposes a framework to develop and ethically evaluate expert-guided multi-modal AI, addressing these challenges in AI integration within precision medicine. We illustrate this framework with case study on insulin management for T1D. To ensure ethical considerations and clinician engagement, we adopt a co-design approach where AI serves an assistive role, with final diagnoses or treatment plans emerging from collaboration between clinicians and AI.

Ayan Banerjee, Sandeep K. S. Gupta

Recovering a physics-driven model, i.e. a governing set of equations of the underlying dynamical systems, from the real-world data has been of recent interest. Most existing methods either operate on simulation data with unrealistically high sampling rates or require explicit measurements of all system variables, which is not amenable in real-world deployments. Moreover, they assume the timestamps of external perturbations to the physical system are known a priori, without uncertainty, implicitly discounting any sensor time-synchronization or human reporting errors. In this paper, we propose a novel liquid time constant neural network (LTC-NN) based architecture to recover underlying model of physical dynamics from real-world data. The automatic differentiation property of LTC-NN nodes overcomes problems associated with low sampling rates, the input dependent time constant in the forward pass of the hidden layer of LTC-NN nodes creates a massive search space of implicit physical dynamics, the physics model solver based data reconstruction loss guides the search for the correct set of implicit dynamics, and the use of the dropout regularization in the dense layer ensures extraction of the sparsest model. Further, to account for the perturbation timing error, we utilize dense layer nodes to search through input shifts that results in the lowest reconstruction loss. Experiments on four benchmark dynamical systems, three with simulation data and one with the real-world data show that the LTC-NN architecture is more accurate in recovering implicit physics model coefficients than the state-of-the-art sparse model recovery approaches. We also introduce four additional case studies (total eight) on real-life medical examples in simulation and with real-world clinical data to show effectiveness of our approach in recovering underlying model in practice.

Payal Kamboj, Ayan Banerjee, Sandeep K. S. Gupta

In domains such as biomedical, expert insights are crucial for selecting the most informative modalities for artificial intelligence (AI) methodologies. However, using all available modalities poses challenges, particularly in determining the impact of each modality on performance and optimizing their combinations for accurate classification. Traditional approaches resort to manual trial and error methods, lacking systematic frameworks for discerning the most relevant modalities. Moreover, although multi-modal learning enables the integration of information from diverse sources, utilizing all available modalities is often impractical and unnecessary. To address this, we introduce an entropy-based algorithm STORM to solve the modality selection problem for rare event. This algorithm systematically evaluates the information content of individual modalities and their combinations, identifying the most discriminative features essential for rare class classification tasks. Through seizure onset zone detection case study, we demonstrate the efficacy of our algorithm in enhancing classification performance. By selecting useful subset of modalities, our approach paves the way for more efficient AI-driven biomedical analyses, thereby advancing disease diagnosis in clinical settings.

Bin Xu, Ayan Banerjee, Midhat Urooj et al.

Digital twins (DTs) can enable precision healthcare by continually learning a mathematical representation of patient-specific dynamics. However, mission critical healthcare applications require fast, resource-efficient DT learning, which is often infeasible with existing model recovery (MR) techniques due to their reliance on iterative solvers and high compute/memory demands. In this paper, we present a general DT learning framework that is amenable to acceleration on reconfigurable hardware such as FPGAs, enabling substantial speedup and energy efficiency. We compare our FPGA-based implementation with a multi-processing implementation in mobile GPU, which is a popular choice for AI in edge devices. Further, we compare both edge AI implementations with cloud GPU baseline. Specifically, our FPGA implementation achieves an 8.8x improvement in \text{performance-per-watt} for the MR task, a 28.5x reduction in DRAM footprint, and a 1.67x runtime speedup compared to cloud GPU baselines. On the other hand, mobile GPU achieves 2x better performance per watts but has 2x increase in runtime and 10x more DRAM footprint than FPGA. We show the usage of this technique in DT guided synthetic data generation for Type 1 Diabetes and proactive coronary artery disease detection.

Ayan Banerjee, Kuntal Thakur, Sandeep Gupta

Generalizing image classification across domains remains challenging in critical tasks such as fundus image-based diabetic retinopathy (DR) grading and resting-state fMRI seizure onset zone (SOZ) detection. When domains differ in unknown causal factors, achieving cross-domain generalization is difficult, and there is no established methodology to objectively assess such differences without direct metadata or protocol-level information from data collectors, which is typically inaccessible. We first introduce domain conformal bounds (DCB), a theoretical framework to evaluate whether domains diverge in unknown causal factors. Building on this, we propose GenEval, a multimodal Vision Language Models (VLM) approach that combines foundational models (e.g., MedGemma-4B) with human knowledge via Low-Rank Adaptation (LoRA) to bridge causal gaps and enhance single-source domain generalization (SDG). Across eight DR and two SOZ datasets, GenEval achieves superior SDG performance, with average accuracy of 69.2% (DR) and 81% (SOZ), outperforming the strongest baselines by 9.4% and 1.8%, respectively.

Bin Xu, Ayan Banerjee, Sandeep Gupta

Model Recovery (MR) is a core primitive for physical AI and real-time digital twins, but GPUs often execute MR inefficiently due to iterative dependencies, kernel-launch overheads, underutilized memory bandwidth, and high data-movement latency. We present MERINDA, an FPGA-accelerated MR framework that restructures computation as a streaming dataflow pipeline. MERINDA exploits on-chip locality through BRAM tiling, fixed-point kernels, and the concurrent use of LUT fabric and carry-chain adders to expose fine-grained spatial parallelism while minimizing off-chip traffic. This hardware-aware formulation removes synchronization bottlenecks and sustains high throughput across the iterative updates in MR. On representative MR workloads, MERINDA delivers up to 6.3x fewer cycles than an FPGA-based LTC baseline, enabling real-time performance for time-critical physical systems.

Ayan Banerjee, Josep Lladós, Umapada Pal et al.

Text-to-story visualization is challenging due to the need for consistent interaction among multiple characters across frames. Existing methods struggle with character consistency, leading to artifact generation and inaccurate dialogue rendering, which results in disjointed storytelling. In response, we introduce TaleDiffusion, a novel framework for generating multi-character stories with an iterative process, maintaining character consistency, and accurate dialogue assignment via postprocessing. Given a story, we use a pre-trained LLM to generate per-frame descriptions, character details, and dialogues via in-context learning, followed by a bounded attention-based per-box mask technique to control character interactions and minimize artifacts. We then apply an identity-consistent self-attention mechanism to ensure character consistency across frames and region-aware cross-attention for precise object placement. Dialogues are also rendered as bubbles and assigned to characters via CLIPSeg. Experimental results demonstrate that TaleDiffusion outperforms existing methods in consistency, noise reduction, and dialogue rendering.

Midhat Urooj, Ayan Banerjee, Farhat Shaikh et al.

Domain generalization remains a critical challenge in medical imaging, where models trained on single sources often fail under real-world distribution shifts. We propose KG-DG, a neuro-symbolic framework for diabetic retinopathy (DR) classification that integrates vision transformers with expert-guided symbolic reasoning to enable robust generalization across unseen domains. Our approach leverages clinical lesion ontologies through structured, rule-based features and retinal vessel segmentation, fusing them with deep visual representations via a confidence-weighted integration strategy. The framework addresses both single-domain generalization (SDG) and multi-domain generalization (MDG) by minimizing the KL divergence between domain embeddings, thereby enforcing alignment of high-level clinical semantics. Extensive experiments across four public datasets (APTOS, EyePACS, Messidor-1, Messidor-2) demonstrate significant improvements: up to a 5.2% accuracy gain in cross-domain settings and a 6% improvement over baseline ViT models. Notably, our symbolic-only model achieves a 63.67% average accuracy in MDG, while the complete neuro-symbolic integration achieves the highest accuracy compared to existing published baselines and benchmarks in challenging SDG scenarios. Ablation studies reveal that lesion-based features (84.65% accuracy) substantially outperform purely neural approaches, confirming that symbolic components act as effective regularizers beyond merely enhancing interpretability. Our findings establish neuro-symbolic integration as a promising paradigm for building clinically robust, and domain-invariant medical AI systems.

Ayan Banerjee, Fernando Vilariño, Josep Lladós

Preserving facial identity under extreme stylistic transformation remains a major challenge in generative art. In graffiti, a high-contrast, abstract medium, subtle distortions to the eyes, nose, or mouth can erase the subject's recognizability, undermining both personal and cultural authenticity. We present CraftGraffiti, an end-to-end text-guided graffiti generation framework designed with facial feature preservation as a primary objective. Given an input image and a style and pose descriptive prompt, CraftGraffiti first applies graffiti style transfer via LoRA-fine-tuned pretrained diffusion transformer, then enforces identity fidelity through a face-consistent self-attention mechanism that augments attention layers with explicit identity embeddings. Pose customization is achieved without keypoints, using CLIP-guided prompt extension to enable dynamic re-posing while retaining facial coherence. We formally justify and empirically validate the "style-first, identity-after" paradigm, showing it reduces attribute drift compared to the reverse order. Quantitative results demonstrate competitive facial feature consistency and state-of-the-art aesthetic and human preference scores, while qualitative analyses and a live deployment at the Cruilla Festival highlight the system's real-world creative impact. CraftGraffiti advances the goal of identity-respectful AI-assisted artistry, offering a principled approach for blending stylistic freedom with recognizability in creative AI applications.

Anindya Mondal, Ayan Banerjee, Sauradip Nag et al.

Diffusion models have shown remarkable progress in photorealistic image synthesis, yet they remain unreliable for generating scenes with a precise number of object instances, particularly in complex and high-density settings. We present CountLoop, a training-free framework that provides diffusion models with accurate instance control through iterative structured feedback. The approach alternates between image generation and multimodal agent evaluation, where a language-guided planner and critic assess object counts, spatial arrangements, and attribute consistency. This feedback is then used to refine layouts and guide subsequent generations. To further improve separation between objects, especially in occluded scenes, we introduce instance-driven attention masking and compositional generation techniques. Experiments on COCO Count, T2I CompBench, and two new high-instance benchmarks show that CountLoop achieves counting accuracy of up to 98% while maintaining spatial fidelity and visual quality, outperforming layout-based and gradient-guided baselines with a score of 0.97.

Payal Kamboj, Ayan Banerjee, Bin Xu et al.

Rare events, due to their infrequent occurrences, do not have much data, and hence deep learning techniques fail in estimating the distribution for such data. Open-vocabulary models represent an innovative approach to image classification. Unlike traditional models, these models classify images into any set of categories specified with natural language prompts during inference. These prompts usually comprise manually crafted templates (e.g., 'a photo of a {}') that are filled in with the names of each category. This paper introduces a simple yet effective method for generating highly accurate and contextually descriptive prompts containing discriminative characteristics. Rare event detection, especially in medicine, is more challenging due to low inter-class and high intra-class variability. To address these, we propose a novel approach that uses domain-specific expert knowledge on rare events to generate customized and contextually relevant prompts, which are then used by large language models for image classification. Our zero-shot, privacy-preserving method enhances rare event classification without additional training, outperforming state-of-the-art techniques.

Jordy Van Landeghem, Subhajit Maity, Ayan Banerjee et al.

This work explores knowledge distillation (KD) for visually-rich document (VRD) applications such as document layout analysis (DLA) and document image classification (DIC). While VRD research is dependent on increasingly sophisticated and cumbersome models, the field has neglected to study efficiency via model compression. Here, we design a KD experimentation methodology for more lean, performant models on document understanding (DU) tasks that are integral within larger task pipelines. We carefully selected KD strategies (response-based, feature-based) for distilling knowledge to and from backbones with different architectures (ResNet, ViT, DiT) and capacities (base, small, tiny). We study what affects the teacher-student knowledge gap and find that some methods (tuned vanilla KD, MSE, SimKD with an apt projector) can consistently outperform supervised student training. Furthermore, we design downstream task setups to evaluate covariate shift and the robustness of distilled DLA models on zero-shot layout-aware document visual question answering (DocVQA). DLA-KD experiments result in a large mAP knowledge gap, which unpredictably translates to downstream robustness, accentuating the need to further explore how to efficiently obtain more semantic document layout awareness.

Sanket Biswas, Ayan Banerjee, Josep Lladós et al.

Understanding documents with rich layouts is an essential step towards information extraction. Business intelligence processes often require the extraction of useful semantic content from documents at a large scale for subsequent decision-making tasks. In this context, instance-level segmentation of different document objects (title, sections, figures etc.) has emerged as an interesting problem for the document analysis and understanding community. To advance the research in this direction, we present a transformer-based model called \emph{DocSegTr} for end-to-end instance segmentation of complex layouts in document images. The method adapts a twin attention module, for semantic reasoning, which helps to become highly computationally efficient compared with the state-of-the-art. To the best of our knowledge, this is the first work on transformer-based document segmentation. Extensive experimentation on competitive benchmarks like PubLayNet, PRIMA, Historical Japanese (HJ) and TableBank demonstrate that our model achieved comparable or better segmentation performance than the existing state-of-the-art approaches with the average precision of 89.4, 40.3, 83.4 and 93.3. This simple and flexible framework could serve as a promising baseline for instance-level recognition tasks in document images.