Bangla-Bayanno: A 52K-Pair Bengali Visual Question Answering Dataset with LLM-Assisted Translation Refinement

In this paper, we introduce Bangla-Bayanno, an open-ended Visual Question Answering (VQA) Dataset in Bangla, a widely used, low-resource language in multimodal AI research. The majority of existing datasets are either manually annotated with an emphasis on a specific domain, query type, or answer type or are constrained by niche answer formats. In order to mitigate human-induced errors and guarantee lucidity, we implemented a multilingual LLM-assisted translation refinement pipeline. This dataset overcomes the issues of low-quality translations from multilingual sources. The dataset comprises 52,650 question-answer pairs across 4750+ images. Questions are classified into three distinct answer types: nominal (short descriptive), quantitative (numeric), and polar (yes/no). Bangla-Bayanno provides the most comprehensive open-source, high-quality VQA benchmark in Bangla, aiming to advance research in low-resource multimodal learning and facilitate the development of more inclusive AI systems.

GeoAware-VLA: Implicit Geometry Aware Vision-Language-Action Model

Vision-Language-Action (VLA) models often fail to generalize to novel camera viewpoints, a limitation stemming from their difficulty in inferring robust 3D geometry from 2D images. We introduce GeoAware-VLA, a simple yet effective approach that enhances viewpoint invariance by integrating strong geometric priors into the vision backbone. Instead of training a visual encoder or relying on explicit 3D data, we leverage a frozen, pretrained geometric vision model as a feature extractor. A trainable projection layer then adapts these geometrically-rich features for the policy decoder, relieving it of the burden of learning 3D consistency from scratch. Through extensive evaluations on LIBERO benchmark subsets, we show GeoAware-VLA achieves substantial improvements in zero-shot generalization to novel camera poses, boosting success rates by over 2x in simulation. Crucially, these benefits translate to the physical world; our model shows a significant performance gain on a real robot, especially when evaluated from unseen camera angles. Our approach proves effective across both continuous and discrete action spaces, highlighting that robust geometric grounding is a key component for creating more generalizable robotic agents.

CoSwin: Convolution Enhanced Hierarchical Shifted Window Attention For Small-Scale Vision

Vision Transformers (ViTs) have achieved impressive results in computer vision by leveraging self-attention to model long-range dependencies. However, their emphasis on global context often comes at the expense of local feature extraction in small datasets, particularly due to the lack of key inductive biases such as locality and translation equivariance. To mitigate this, we propose CoSwin, a novel feature-fusion architecture that augments the hierarchical shifted window attention with localized convolutional feature learning. Specifically, CoSwin integrates a learnable local feature enhancement module into each attention block, enabling the model to simultaneously capture fine-grained spatial details and global semantic structure. We evaluate CoSwin on multiple image classification benchmarks including CIFAR-10, CIFAR-100, MNIST, SVHN, and Tiny ImageNet. Our experimental results show consistent performance gains over state-of-the-art convolutional and transformer-based models. Notably, CoSwin achieves improvements of 2.17% on CIFAR-10, 4.92% on CIFAR-100, 0.10% on MNIST, 0.26% on SVHN, and 4.47% on Tiny ImageNet over the baseline Swin Transformer. These improvements underscore the effectiveness of local-global feature fusion in enhancing the generalization and robustness of transformers for small-scale vision. Code and pretrained weights available at https://github.com/puskal-khadka/coswin

Integrating Anatomical Priors into a Causal Diffusion Model

3D brain MRI studies often examine subtle morphometric differences between cohorts that are hard to detect visually. Given the high cost of MRI acquisition, these studies could greatly benefit from image syntheses, particularly counterfactual image generation, as seen in other domains, such as computer vision. However, counterfactual models struggle to produce anatomically plausible MRIs due to the lack of explicit inductive biases to preserve fine-grained anatomical details. This shortcoming arises from the training of the models aiming to optimize for the overall appearance of the images (e.g., via cross-entropy) rather than preserving subtle, yet medically relevant, local variations across subjects. To preserve subtle variations, we propose to explicitly integrate anatomical constraints on a voxel-level as prior into a generative diffusion framework. Called Probabilistic Causal Graph Model (PCGM), the approach captures anatomical constraints via a probabilistic graph module and translates those constraints into spatial binary masks of regions where subtle variations occur. The masks (encoded by a 3D extension of ControlNet) constrain a novel counterfactual denoising UNet, whose encodings are then transferred into high-quality brain MRIs via our 3D diffusion decoder. Extensive experiments on multiple datasets demonstrate that PCGM generates structural brain MRIs of higher quality than several baseline approaches. Furthermore, we show for the first time that brain measurements extracted from counterfactuals (generated by PCGM) replicate the subtle effects of a disease on cortical brain regions previously reported in the neuroscience literature. This achievement is an important milestone in the use of synthetic MRIs in studies investigating subtle morphological differences.

Camera Pose Refinement via 3D Gaussian Splatting

Camera pose refinement aims at improving the accuracy of initial pose estimation for applications in 3D computer vision. Most refinement approaches rely on 2D-3D correspondences with specific descriptors or dedicated networks, requiring reconstructing the scene again for a different descriptor or fully retraining the network for each scene. Some recent methods instead infer pose from feature similarity, but their lack of geometry constraints results in less accuracy. To overcome these limitations, we propose a novel camera pose refinement framework leveraging 3D Gaussian Splatting (3DGS), referred to as GS-SMC. Given the widespread usage of 3DGS, our method can employ an existing 3DGS model to render novel views, providing a lightweight solution that can be directly applied to diverse scenes without additional training or fine-tuning. Specifically, we introduce an iterative optimization approach, which refines the camera pose using epipolar geometric constraints among the query and multiple rendered images. Our method allows flexibly choosing feature extractors and matchers to establish these constraints. Extensive empirical evaluations on the 7-Scenes and the Cambridge Landmarks datasets demonstrate that our method outperforms state-of-the-art camera pose refinement approaches, achieving 53.3% and 56.9% reductions in median translation and rotation errors on 7-Scenes, and 40.7% and 53.2% on Cambridge.

Beyond flattening: a geometrically principled positional encoding for vision transformers with Weierstrass elliptic functions

Vision Transformers have demonstrated remarkable success in computer vision tasks, yet their reliance on learnable one-dimensional positional embeddings fundamentally disrupts the inherent two-dimensional spatial structure of images through patch flattening procedures. Traditional positional encoding approaches lack geometric constraints and fail to establish monotonic correspondence between Euclidean spatial distances and sequential index distances, thereby limiting the model's capacity to leverage spatial proximity priors effectively. We propose Weierstrass Elliptic Function Positional Encoding (WEF-PE), a mathematically principled approach that directly addresses two-dimensional coordinates through natural complex domain representation, where the doubly periodic properties of elliptic functions align remarkably with translational invariance patterns commonly observed in visual data. Our method exploits the non-linear geometric nature of elliptic functions to encode spatial distance relationships naturally, while the algebraic addition formula enables direct derivation of relative positional information between arbitrary patch pairs from their absolute encodings. Comprehensive experiments demonstrate that WEF-PE achieves superior performance across diverse scenarios, including 63.78\% accuracy on CIFAR-100 from-scratch training with ViT-Tiny architecture, 93.28\% on CIFAR-100 fine-tuning with ViT-Base, and consistent improvements on VTAB-1k benchmark tasks. Theoretical analysis confirms the distance-decay property through rigorous mathematical proof, while attention visualization reveals enhanced geometric inductive bias and more coherent semantic focus compared to conventional approaches.The source code implementing the methods described in this paper is publicly available on GitHub.

HERMES: Human-to-Robot Embodied Learning from Multi-Source Motion Data for Mobile Dexterous Manipulation

Leveraging human motion data to impart robots with versatile manipulation skills has emerged as a promising paradigm in robotic manipulation. Nevertheless, translating multi-source human hand motions into feasible robot behaviors remains challenging, particularly for robots equipped with multi-fingered dexterous hands characterized by complex, high-dimensional action spaces. Moreover, existing approaches often struggle to produce policies capable of adapting to diverse environmental conditions. In this paper, we introduce HERMES, a human-to-robot learning framework for mobile bimanual dexterous manipulation. First, HERMES formulates a unified reinforcement learning approach capable of seamlessly transforming heterogeneous human hand motions from multiple sources into physically plausible robotic behaviors. Subsequently, to mitigate the sim2real gap, we devise an end-to-end, depth image-based sim2real transfer method for improved generalization to real-world scenarios. Furthermore, to enable autonomous operation in varied and unstructured environments, we augment the navigation foundation model with a closed-loop Perspective-n-Point (PnP) localization mechanism, ensuring precise alignment of visual goals and effectively bridging autonomous navigation and dexterous manipulation. Extensive experimental results demonstrate that HERMES consistently exhibits generalizable behaviors across diverse, in-the-wild scenarios, successfully performing numerous complex mobile bimanual dexterous manipulation tasks. Project Page:https://gemcollector.github.io/HERMES/.

SATURN: Autoregressive Image Generation Guided by Scene Graphs

State-of-the-art text-to-image models excel at photorealistic rendering but often struggle to capture the layout and object relationships implied by complex prompts. Scene graphs provide a natural structural prior, yet previous graph-guided approaches have typically relied on heavy GAN or diffusion pipelines, which lag behind modern autoregressive architectures in both speed and fidelity. We introduce SATURN (Structured Arrangement of Triplets for Unified Rendering Networks), a lightweight extension to VAR-CLIP that translates a scene graph into a salience-ordered token sequence, enabling a frozen CLIP-VQ-VAE backbone to interpret graph structure while fine-tuning only the VAR transformer. On the Visual Genome dataset, SATURN reduces FID from 56.45% to 21.62% and increases the Inception Score from 16.03 to 24.78, outperforming prior methods such as SG2IM and SGDiff without requiring extra modules or multi-stage training. Qualitative results further confirm improvements in object count fidelity and spatial relation accuracy, showing that SATURN effectively combines structural awareness with state-of-the-art autoregressive fidelity.

Designing Practical Models for Isolated Word Visual Speech Recognition

Visual speech recognition (VSR) systems decode spoken words from an input sequence using only the video data. Practical applications of such systems include medical assistance as well as human-machine interactions. A VSR system is typically employed in a complementary role in cases where the audio is corrupt or not available. In order to accurately predict the spoken words, these architectures often rely on deep neural networks in order to extract meaningful representations from the input sequence. While deep architectures achieve impressive recognition performance, relying on such models incurs significant computation costs which translates into increased resource demands in terms of hardware requirements and results in limited applicability in real-world scenarios where resources might be constrained. This factor prevents wider adoption and deployment of speech recognition systems in more practical applications. In this work, we aim to alleviate this issue by developing architectures for VSR that have low hardware costs. Following the standard two-network design paradigm, where one network handles visual feature extraction and another one utilizes the extracted features to classify the entire sequence, we develop lightweight end-to-end architectures by first benchmarking efficient models from the image classification literature, and then adopting lightweight block designs in a temporal convolution network backbone. We create several unified models with low resource requirements but strong recognition performance. Experiments on the largest public database for English words demonstrate the effectiveness and practicality of our developed models. Code and trained models will be made publicly available.

Why Stop at Words? Unveiling the Bigger Picture through Line-Level OCR

Conventional optical character recognition (OCR) techniques segmented each character and then recognized. This made them prone to error in character segmentation, and devoid of context to exploit language models. Advances in sequence to sequence translation in last decade led to modern techniques first detecting words and then inputting one word at a time to a model to directly output full words as sequence of characters. This allowed better utilization of language models and bypass error-prone character segmentation step. We observe that the above transition in style has moved the bottleneck in accuracy to word segmentation. Hence, in this paper, we propose a natural and logical progression from word level OCR to line-level OCR. The proposal allows to bypass errors in word detection, and provides larger sentence context for better utilization of language models. We show that the proposed technique not only improves the accuracy but also efficiency of OCR. Despite our thorough literature survey, we did not find any public dataset to train and benchmark such shift from word to line-level OCR. Hence, we also contribute a meticulously curated dataset of 251 English page images with line-level annotations. Our experimentation revealed a notable end-to-end accuracy improvement of 5.4%, underscoring the potential benefits of transitioning towards line-level OCR, especially for document images. We also report a 4 times improvement in efficiency compared to word-based pipelines. With continuous improvements in large language models, our methodology also holds potential to exploit such advances. Project Website: https://nishitanand.github.io/line-level-ocr-website