Sanyam Jain

Mannat Khurana, Sanyam Jain, Rishav Agarwal

Animation elevates digital documents into immersive experiences, yet creating custom motion paths remains cumbersome, requiring designers to manually select presets, plot Bézier points, and configure timing properties. We introduce Generative Animations, a system that transforms natural language prompts into production-ready animations. By chaining Large Language Models (LLMs) for semantic parsing with the Segment Anything Model (SAM) for visual grounding, our pipeline automatically generates motion paths that respect scene geometry, handle depth-based occlusions, and honor 3D perspective transforms. We demonstrate the system through three use cases: contour-following trajectories, orbital animations with z-order awareness, and perspective-aligned motion on transformed objects.

Yashwardhan Chaudhuri, Sanyam Jain, Paridhi Mundra

We present E3, an automated review assistant that augments reviewers and engineering teams by identifying decision-relevant technical concerns in research papers. For each concern, E3 reports its nature, its location, its bearing on the contribution, and the analysis or evidence that would resolve it, covering unsupported claims, missing ablations, weak baselines, hidden assumptions, threats to validity, and leakage risks. To evaluate E3 without contamination confounds we adopt an issue-level backtesting protocol: the corpus is restricted to papers postdating the training cutoff of every automated source, and for each paper a meta-judge that observes only anonymised reviews labels every issue-source pair as Caught, Partial, or Missed. Applied to 100 ICLR 2026 papers and 4598 judged issue rows, comparing E3 against the ICLR human reviews and two prompt-matched LLM baselines built on gpt-5.4 from OpenAI and claude-opus-4-6 from Anthropic, with meta-judge gpt-5.5, E3 attains the highest recall on every aggregate metric. Partial-inclusive recall reaches 90.2 percent, which is 15.5 points over GPT, 17.1 points over Claude, and 29.2 points over the human reviews, and strict recall preserves the ordering at 65.8 percent. On concerns raised by the human reviewers, E3 recovers 89.6 percent; on concerns the human reviewers missed it surfaces 1635 additional rows admitted into the judged union, 406 above the next-best source. Corpus, baseline prompts, judge prompt template, and evaluation code are released.

Johan Andreas Balle Rubak, Sara Haghighat, Sanyam Jain et al.

Impaction of the mandibular third molar in proximity to the mandibular canal increases the risk of inferior alveolar nerve injury. Panoramic radiography is routinely used to assess this relationship. Automated classification of molar-canal overlap could support clinical triage and reduce unnecessary CBCT referrals, while federated learning (FL) enables multi-center collaboration without sharing patient data. We compared Local Learning (LL), FL, and Centralized Learning (CL) for binary overlap/no-overlap classification on cropped panoramic radiographs partitioned across eight independent labelers. A pretrained ResNet-34 was trained under each paradigm and evaluated using per-client metrics with locally optimized thresholds and pooled test performance with a global threshold. Performance was assessed using area under the receiver operating characteristic curve (AUC) and threshold-based metrics, alongside training dynamics, Grad-CAM visualizations, and server-side aggregate monitoring signals. On the test set, CL achieved the highest performance (AUC 0.831; accuracy = 0.782), FL showed intermediate performance (AUC 0.757; accuracy = 0.703), and LL generalized poorly across clients (AUC range = 0.619-0.734; mean = 0.672). Training curves suggested overfitting, particularly in LL models, and Grad-CAM indicated more anatomically focused attention in CL and FL. Overall, centralized training provided the strongest performance, while FL offers a privacy-preserving alternative that outperforms LL.

N. A. Adarsh Pritam, Jeba Shiney O, Sanyam Jain

This work introduces SkinGenBench, a systematic biomedical imaging benchmark that investigates how preprocessing complexity interacts with generative model choice for synthetic dermoscopic image augmentation and downstream melanoma diagnosis. Using a curated dataset of 14,116 dermoscopic images from HAM10000 and MILK10K across five lesion classes, we evaluate the two representative generative paradigms: StyleGAN2-ADA and Denoising Diffusion Probabilistic Models (DDPMs) under basic geometric augmentation and advanced artifact removal pipelines. Synthetic melanoma images are assessed using established perceptual and distributional metrics (FID, KID, IS), feature space analysis, and their impact on diagnostic performance across five downstream classifiers. Experimental results demonstrate that generative architecture choice has a stronger influence on both image fidelity and diagnostic utility than preprocessing complexity. StyleGAN2-ADA consistently produced synthetic images more closely aligned with real data distributions, achieving the lowest FID (~65.5) and KID (~0.05), while diffusion models generated higher variance samples at the cost of reduces perceptual fidelity and class anchoring. Advanced artifact removal yielded only marginal improvements in generative metrics and provided limited downstream diagnostic gains, suggesting possible suppression of clinically relevant texture cues. In contrast, synthetic data augmentation substantially improved melanoma detection with 8-15% absolute gains in melanoma F1-score, and ViT-B/16 achieving F1~0.88 and ROC-AUC~0.98, representing an improvement of approximately 14% over non-augmented baselines. Our code can be found at https://github.com/adarsh-crafts/SkinGenBench

Sanyam Jain

For a long time, detecting hand gestures and recognizing them as letters or numbers has been a challenging task. This creates communication barriers for individuals with disabilities. This paper introduces a new dataset, the Annotated Dataset for Danish Sign Language (ADDSL). Annota-tions for the dataset were made using the open-source tool LabelImg in the YOLO format. Using this dataset, a one-stage ob-ject detector model (YOLOv5) was trained with the CSP-DarkNet53 backbone and YOLOv3 head to recognize letters (A-Z) and numbers (0-9) using only seven unique images per class (without augmen-tation). Five models were trained with 350 epochs, resulting in an average inference time of 9.02ms per image and a best accu-racy of 92% when compared to previous research. Our results show that modified model is efficient and more accurate than existing work in the same field. The code repository for our model is available at the GitHub repository https://github.com/s4nyam/pvt-addsl.

Sanyam Jain

This research paper presents an experimental approach to using the Reptile algorithm for reinforcement learning to train a neural network to play Super Mario Bros. We implement the Reptile algorithm using the Super Mario Bros Gym library and TensorFlow in Python, creating a neural network model with a single convolutional layer, a flatten layer, and a dense layer. We define the optimizer and use the Reptile class to create an instance of the Reptile meta-learning algorithm. We train the model using multiple tasks and episodes, choosing actions using the current weights of the neural network model, taking those actions in the environment, and updating the model weights using the Reptile algorithm. We evaluate the performance of the algorithm by printing the total reward for each episode. In addition, we compare the performance of the Reptile algorithm approach to two other popular reinforcement learning algorithms, Proximal Policy Optimization (PPO) and Deep Q-Network (DQN), applied to the same Super Mario Bros task. Our results demonstrate that the Reptile algorithm provides a promising approach to few-shot learning in video game AI, with comparable or even better performance than the other two algorithms, particularly in terms of moves vs distance that agent performs for 1M episodes of training. The results shows that best total distance for world 1-2 in the game environment were ~1732 (PPO), ~1840 (DQN) and ~2300 (RAMario). Full code is available at https://github.com/s4nyam/RAMario.

Deergh Singh Budhauria, Sanyam Jain, Rishav Agarwal et al.

Globalization of graphic designs such as those used in marketing materials and magazines is increasingly important for communication to broad audiences. To accomplish this, the textual content in the graphic designs needs to be accurately translated and have the text styling preserved in order to fit visually into the design. Preserving text styling requires high accuracy word alignment between the original and the translated text. The problem of word alignment between source and translated text is long known. The industry standards for extracting word alignments are defined by Giza++ and attention probabilities from neural machine translation (NMT) models. In this paper, we explore three new methods to tackle the word alignment problem for transferring text styles from the source to the translated text. The proposed methods are developed on top of commercially available NMT and LLM translation technologies. They include: NMT with custom input and output tags for text styling; LLM with custom input and output tags; a hybrid with NMT for translation followed by an LLM with use of unigram mappings. To analyze the performance of these solutions, their alignment results are compared with the results of an attention head approach to gauge their usability in graphic design applications. Interestingly, the attention head strong baseline proves more accurate than the LLM or NMT approach and on par with the hybrid NMT+LLM approach.

Sanyam Jain, Pragya Kandari, Manit Singhal et al.

Generative object compositing methods have shown remarkable ability to seamlessly insert objects into scenes. However, when applied to real-world catalog image generation, these methods require tedious manual intervention: users must carefully adjust masks when product dimensions differ, and painstakingly restore occluded elements post-generation. We present CatalogStitch, a set of model-agnostic techniques that automate these corrections, enabling user-friendly content creation. Our dimension-aware mask computation algorithm automatically adapts the target region to accommodate products with different dimensions; users simply provide a product image and background, without manual mask adjustments. Our occlusion-aware hybrid restoration method guarantees pixel-perfect preservation of occluding elements, eliminating post-editing workflows. We additionally introduce CatalogStitch-Eval, a 58-example benchmark covering aspect-ratio mismatch and occlusion-heavy catalog scenarios, together with supplementary PDF and HTML viewers. We evaluate our techniques with three state-of-the-art compositing models (ObjectStitch, OmniPaint, and InsertAnything), demonstrating consistent improvements across diverse catalog scenarios. By reducing manual intervention and automating tedious corrections, our approach transforms generative compositing into a practical, human-friendly tool for production catalog workflows.

Soren Pedersen, Sanyam Jain, Mikkel Chavez et al.

This paper presents the development of a generative adversarial network (GAN) for synthesizing dental panoramic radiographs. Although exploratory in nature, the study aims to address the scarcity of data in dental research and education. We trained a deep convolutional GAN (DCGAN) using a Wasserstein loss with gradient penalty (WGANGP) on a dataset of 2322 radiographs of varying quality. The focus was on the dentoalveolar regions, other anatomical structures were cropped out. Extensive preprocessing and data cleaning were performed to standardize the inputs while preserving anatomical variability. We explored four candidate models by varying critic iterations, feature depth, and the use of denoising prior to training. A clinical expert evaluated the generated radiographs based on anatomical visibility and realism, using a 5-point scale (1 very poor 5 excellent). Most images showed moderate anatomical depiction, although some were degraded by artifacts. A trade-off was observed the model trained on non-denoised data yielded finer details especially in structures like the mandibular canal and trabecular bone, while a model trained on denoised data offered superior overall image clarity and sharpness. These findings provide a foundation for future work on GAN-based methods in dental imaging.

Johan Andreas Balle Rubak, Khuram Naveed, Sanyam Jain et al.

Objectives: Federated learning (FL) may mitigate privacy constraints, heterogeneous data quality, and inconsistent labeling in dental diagnostic AI. We compared FL with centralized (CL) and local learning (LL) for tooth segmentation in panoramic radiographs across multiple data corruption scenarios. Methods: An Attention U-Net was trained on 2066 radiographs from six institutions across four settings: baseline (unaltered data); label manipulation (dilated/missing annotations); image-quality manipulation (additive Gaussian noise); and exclusion of a faulty client with corrupted data. FL was implemented via the Flower AI framework. Per-client training- and validation-loss trajectories were monitored for anomaly detection and a set of metrics (Dice, IoU, HD, HD95 and ASSD) was evaluated on a hold-out test set. From these metrics significance results were reported through Wilcoxon signed-rank test. CL and LL served as comparators. Results: Baseline: FL achieved a median Dice of 0.94889 (ASSD: 1.33229), slightly better than CL at 0.94706 (ASSD: 1.37074) and LL at 0.93557-0.94026 (ASSD: 1.51910-1.69777). Label manipulation: FL maintained the best median Dice score at 0.94884 (ASSD: 1.46487) versus CL's 0.94183 (ASSD: 1.75738) and LL's 0.93003-0.94026 (ASSD: 1.51910-2.11462). Image noise: FL led with Dice at 0.94853 (ASSD: 1.31088); CL scored 0.94787 (ASSD: 1.36131); LL ranged from 0.93179-0.94026 (ASSD: 1.51910-1.77350). Faulty-client exclusion: FL reached Dice at 0.94790 (ASSD: 1.33113) better than CL's 0.94550 (ASSD: 1.39318). Loss-curve monitoring reliably flagged the corrupted site. Conclusions: FL matches or exceeds CL and outperforms LL across corruption scenarios while preserving privacy. Per-client loss trajectories provide an effective anomaly-detection mechanism and support FL as a practical, privacy-preserving approach for scalable clinical AI deployment.

Sanyam Jain, Khuram Naveed, Illia Oleksiienko et al.

This work introduces InJecteD, a framework for interpreting Denoising Diffusion Probabilistic Models (DDPMs) by analyzing sample trajectories during the denoising process of 2D point cloud generation. We apply this framework to three datasets from the Datasaurus Dozen bullseye, dino, and circle using a simplified DDPM architecture with customizable input and time embeddings. Our approach quantifies trajectory properties, including displacement, velocity, clustering, and drift field dynamics, using statistical metrics such as Wasserstein distance and cosine similarity. By enhancing model transparency, InJecteD supports human AI collaboration by enabling practitioners to debug and refine generative models. Experiments reveal distinct denoising phases: initial noise exploration, rapid shape formation, and final refinement, with dataset-specific behaviors example, bullseyes concentric convergence vs. dinos complex contour formation. We evaluate four model configurations, varying embeddings and noise schedules, demonstrating that Fourier based embeddings improve trajectory stability and reconstruction quality

Utkarsh Shandilya, Marsha Mariya Kappan, Sanyam Jain et al.

Human action recognition plays a critical role in healthcare and medicine, supporting applications such as patient behavior monitoring, fall detection, surgical robot supervision, and procedural skill assessment. While traditional models like CNNs and RNNs have achieved moderate success, they often struggle to generalize across diverse and complex actions. Recent advancements in vision-language models, especially the transformer-based CLIP model, offer promising capabilities for generalizing action recognition from video data. In this work, we evaluate CLIP on the UCF-101 dataset and systematically analyze its performance under three masking strategies: (1) percentage-based and shape-based black masking at 10%, 30%, and 50%, (2) feature-specific masking to suppress bias-inducing elements, and (3) isolation masking that retains only class-specific regions. Our results reveal that CLIP exhibits inconsistent behavior and frequent misclassifications, particularly when essential visual cues are obscured. To overcome these limitations, we propose incorporating class-specific noise, learned via a custom loss function, to reinforce attention to class-defining features. This enhancement improves classification accuracy and model confidence while reducing bias. We conclude with a discussion on the challenges of applying such models in clinical domains and outline directions for future work to improve generalizability across domain-independent healthcare scenarios.

Sanyam Jain, Bruna Neves de Freitas, Andreas Basse-OConnor et al.

There has been increasing interest in the generation of high-quality, realistic synthetic medical images in recent years. Such synthetic datasets can mitigate the scarcity of public datasets for artificial intelligence research, and can also be used for educational purposes. In this paper, we propose a combination of diffusion-based generation (PanoDiff) and Super-Resolution (SR) for generating synthetic dental panoramic radiographs (PRs). The former generates a low-resolution (LR) seed of a PR (256 X 128) which is then processed by the SR model to yield a high-resolution (HR) PR of size 1024 X 512. For SR, we propose a state-of-the-art transformer that learns local-global relationships, resulting in sharper edges and textures. Experimental results demonstrate a Frechet inception distance score of 40.69 between 7243 real and synthetic images (in HR). Inception scores were 2.55, 2.30, 2.90 and 2.98 for real HR, synthetic HR, real LR and synthetic LR images, respectively. Among a diverse group of six clinical experts, all evaluating a mixture of 100 synthetic and 100 real PRs in a time-limited observation, the average accuracy in distinguishing real from synthetic images was 68.5% (with 50% corresponding to random guessing).

Sanyam Jain

This project proposes and compares a new way to optimise Super Mario Bros. (SMB) environment where the control is in hand of two approaches, namely, Genetic Algorithm (MarioGA) and NeuroEvolution (MarioNE). Not only we learn playing SMB using these techniques, but also optimise it with constrains of collection of coins and finishing levels. Firstly, we formalise the SMB agent to maximize the total value of collected coins (reward) and maximising the total distance traveled (reward) in order to finish the level faster (time penalty) for both the algorithms. Secondly, we study MarioGA and its evaluation function (fitness criteria) including its representation methods, crossover used, mutation operator formalism, selection method used, MarioGA loop, and few other parameters. Thirdly, MarioNE is applied on SMB where a population of ANNs with random weights is generated, and these networks control Marios actions in the game. Fourth, SMB is further constrained to complete the task within the specified time, rebirths (deaths) within the limit, and performs actions or moves within the maximum allowed moves, while seeking to maximize the total coin value collected. This ensures an efficient way of finishing SMB levels. Finally, we provide a fivefold comparative analysis by plotting fitness plots, ability to finish different levels of world 1, and domain adaptation (transfer learning) of the trained models.

Sanyam Jain

This paper implements and investigates popular adversarial attacks on the YOLOv5 Object Detection algorithm. The paper explores the vulnerability of the YOLOv5 to adversarial attacks in the context of traffic and road sign detection. The paper investigates the impact of different types of attacks, including the Limited memory Broyden Fletcher Goldfarb Shanno (L-BFGS), the Fast Gradient Sign Method (FGSM) attack, the Carlini and Wagner (C&W) attack, the Basic Iterative Method (BIM) attack, the Projected Gradient Descent (PGD) attack, One Pixel Attack, and the Universal Adversarial Perturbations attack on the accuracy of YOLOv5 in detecting traffic and road signs. The results show that YOLOv5 is susceptible to these attacks, with misclassification rates increasing as the magnitude of the perturbations increases. We also explain the results using saliency maps. The findings of this paper have important implications for the safety and reliability of object detection algorithms used in traffic and transportation systems, highlighting the need for more robust and secure models to ensure their effectiveness in real-world applications.

Sanyam Jain

Marine animals and deep underwater objects are difficult to recognize and monitor for safety of aquatic life. There is an increasing challenge when the water is saline with granular particles and impurities. In such natural adversarial environment, traditional approaches like CNN start to fail and are expensive to compute. This project involves implementing and evaluating various object detection models, including EfficientDet, YOLOv5, YOLOv8, and Detectron2, on an existing annotated underwater dataset, called the Brackish-Dataset. The dataset comprises annotated image sequences of fish, crabs, starfish, and other aquatic animals captured in Limfjorden water with limited visibility. The aim of this research project is to study the efficiency of newer models on the same dataset and contrast them with the previous results based on accuracy and inference time. Firstly, I compare the results of YOLOv3 (31.10% mean Average Precision (mAP)), YOLOv4 (83.72% mAP), YOLOv5 (97.6%), YOLOv8 (98.20%), EfficientDet (98.56% mAP) and Detectron2 (95.20% mAP) on the same dataset. Secondly, I provide a modified BiSkFPN mechanism (BiFPN neck with skip connections) to perform complex feature fusion in adversarial noise which makes modified EfficientDet robust to perturbations. Third, analyzed the effect on accuracy of EfficientDet (98.63% mAP) and YOLOv5 by adversarial learning (98.04% mAP). Last, I provide class activation map based explanations (CAM) for the two models to promote Explainability in black box models. Overall, the results indicate that modified EfficientDet achieved higher accuracy with five-fold cross validation than the other models with 88.54% IoU of feature maps.

Sanyam Jain

Training machine learning models in an incremental fashion is not only important but also an efficient way to achieve artificial general intelligence. The ability that humans possess of continuous or lifelong learning helps them to not forget previously learned tasks. However, current neural network models are prone to catastrophic forgetting when it comes to continual learning. Many researchers have come up with several techniques in order to reduce the effect of forgetting from neural networks, however, all techniques are studied classically with a very less focus on changing the machine learning model architecture. In this research paper, we show that it is not only possible to circumvent catastrophic forgetting in continual learning with novel hybrid classical-quantum neural networks, but also explains what features are most important to learn for classification. In addition, we also claim that if the model is trained with these explanations, it tends to give better performance and learn specific features that are far from the decision boundary. Finally, we present the experimental results to show comparisons between classical and classical-quantum hybrid architectures on benchmark MNIST and CIFAR-10 datasets. After successful runs of learning procedure, we found hybrid neural network outperforms classical one in terms of remembering the right evidences of the class-specific features.

Sanyam Jain, Aarati Shrestha, Stefano Nichele

This research project investigates Lenia, an artificial life platform that simulates ecosystems of digital creatures. Lenia's ecosystem consists of simple, artificial organisms that can move, consume, grow, and reproduce. The platform is important as a tool for studying artificial life and evolution, as it provides a scalable and flexible environment for creating a diverse range of organisms with varying abilities and behaviors. Measuring complexity in Lenia is a key aspect of the study, which identifies the metrics for measuring long-term complex emerging behavior of rules, with the aim of evolving better Lenia behaviors which are yet not discovered. The Genetic Algorithm uses neighborhoods or kernels as genotype while keeping the rest of the parameters of Lenia as fixed, for example growth function, to produce different behaviors respective to the population and then measures fitness value to decide the complexity of the resulting behavior. First, we use Variation over Time as a fitness function where higher variance between the frames are rewarded. Second, we use Auto-encoder based fitness where variation of the list of reconstruction loss for the frames is rewarded. Third, we perform combined fitness where higher variation of the pixel density of reconstructed frames is rewarded. All three experiments are tweaked with pixel alive threshold and frames used. Finally, after performing nine experiments of each fitness for 500 generations, we pick configurations from all experiments such that there is a scope of further evolution, and run it for 2500 generations. Results show that the kernel's center of mass increases with a specific set of pixels and together with borders the kernel try to achieve a Gaussian distribution. Results are available at https://s4nyam.github.io/evolenia/