Search papers, labs, and topics across Lattice.
100 papers published across 5 labs.
Sparse graph policies not only outperform traditional image-based methods but also expose hidden dataset biases, enhancing both performance and interpretability in robot learning.
NeuWorld achieves impressive long-horizon consistency in interactive video generation without relying on pretrained models or auxiliary 3D reconstruction methods.
Flux-GS slashes the storage overhead for mobile 3D rendering while preserving essential visual fidelity, enabling high-quality graphics on resource-constrained devices.
InnerZoom cuts end-to-end latency by up to 31.8% while surpassing existing GUI grounding methods, proving that less can indeed be more.
Dynamic token editing in image synthesis could redefine how we approach high-resolution generative models.
NeuWorld achieves impressive long-horizon consistency in interactive video generation without relying on pretrained models or auxiliary 3D reconstruction methods.
Flux-GS slashes the storage overhead for mobile 3D rendering while preserving essential visual fidelity, enabling high-quality graphics on resource-constrained devices.
InnerZoom cuts end-to-end latency by up to 31.8% while surpassing existing GUI grounding methods, proving that less can indeed be more.
Dynamic token editing in image synthesis could redefine how we approach high-resolution generative models.
Achieving a 95.8% success rate in resolving self-collisions, PoseShield transforms how we handle human pose estimation under extreme articulations.
Training deepfake detectors solely on real images can lead to significantly better generalization across unseen generator types.
Uncertainty modeling in brain tumor segmentation can significantly enhance reliability, even when critical MRI modalities are missing.
Achieving physically grounded global illumination rendering at scale, RenderFormer++ outperforms prior methods in both accuracy and efficiency.
TRACE redefines glioblastoma response assessment by framing it as structured concept reasoning, significantly enhancing interpretability and clinical relevance.
CVLC achieves a remarkable 16% performance boost in few-shot domain incremental learning, challenging the notion that more data is always necessary for effective adaptation.
Image stimuli consistently reveal robust viewing mode distinctions, while user behavior is intricately tied to specific contexts, showcasing the complexity of human gaze patterns.
Object detection accuracy can now serve as a bridge to translate image dehazing improvements into real-world visibility gains for maritime safety.
False positives in prostate MRI can closely mimic true cancers, complicating diagnosis and treatment decisions.
A novel framework enables zero-shot localization across ground and drone views, outperforming traditional methods by leveraging a rich dataset and advanced geometric modeling.
FFAvatar can reconstruct dynamic, identity-consistent 4D head avatars from just a few images, revolutionizing avatar creation for virtual environments.
Diffusion models not only enhance the accuracy of glaucoma visual field predictions but also provide a robust framework for uncertainty-aware risk assessment in clinical settings.
Low early cue precision can enhance conflict behavior in visual classifiers, but shortcut-rich fine-tuning risks erasing this advantage.
Sparse fusion in RGB-T detection slashes computational costs while preserving accuracy, enabling efficient processing of high-resolution images.
Adaptive imagination can transform limited target data into reliable rollouts, enabling robust sim-to-real transfer in visual reinforcement learning.
A groundbreaking dataset of 470 whole-slide images reveals the potential for AI to transform breast cancer diagnostics through detailed patch-wise classification.
Non-linear neural functional maps can drastically improve 3D shape matching accuracy in the presence of noise and partial data.
IBRSteG enables the undetectable embedding of secret 3D scenes with superior capacity and security, all while eliminating the need for scene-specific optimization.
Stealthy near-infrared attacks can successfully manipulate traffic sign classification, posing significant risks to autonomous vehicle safety.
Achieving a 46.2% reduction in computational cost for CNNs without retraining could revolutionize how we deploy deep learning on edge devices.
A new dataset and transformer model that together nearly double the success rate for 3D hand-object pose estimation in real-world settings.
StereoGS achieves unprecedented accuracy in sparse-view 3D Gaussian Splatting by enforcing binocular consistency, overcoming the limitations of monocular depth priors.
Flood mapping accuracy can be dramatically improved by leveraging a novel cross-sensor learning approach that integrates SAR and optical imagery, achieving AUPRC scores above 0.95.
Learnable latent prompts can stabilize multimodal models and maintain performance even when up to 90% of input modalities are missing.
Achieving robust pose tracking and memory efficiency, KiloGS-SLAM can handle over 10,000 frames in challenging outdoor environments without sacrificing accuracy or detail.
A novel transformer framework achieves superior assessment of rehabilitation exercises by effectively extracting and utilizing joint position features from RGBD data.
Attention redistribution can dramatically enhance facial expression understanding in videos, leading to improved performance on challenging recognition tasks.
CRST boosts retrieval performance in low-resolution surveillance by mitigating the impact of resolution variance, achieving significant improvements without sacrificing high-resolution accuracy.
JPEG AI emerges as the top performer for face recognition in ultra-compressed images, revealing that significant compression can be achieved without sacrificing accuracy.
Goku redefines the landscape of video editing datasets by enabling complex, multi-task editing capabilities that surpass traditional single-task limitations.
A single framework can seamlessly integrate multiple advanced segmentation techniques, revolutionizing how medical image analysis is conducted.
EcoVideo achieves up to 2.9x faster video generation in resource-constrained environments by intelligently orchestrating cloud-edge dynamics.
Achieving precise control over human motion and camera trajectories in 3D scene animation could redefine the realism and interactivity of video generation.
PGE-SAM not only recovers fine details lost to degradation but also adapts to user input, redefining interactive segmentation in challenging imaging conditions.
RBE-Flow outperforms traditional methods by effectively integrating uncertainty into the cross-modal registration process, achieving remarkable accuracy even in the presence of severe radiometric discrepancies.
SA-Homo achieves unprecedented precision in homography estimation, maintaining accuracy even with 8× scale discrepancies that typically cripple existing methods.
Transforming point seeds into temporally consistent object instances, PS-Track sets a new benchmark in multi-object tracking without relying on traditional bounding box annotations.
Sparse graph policies not only outperform traditional image-based methods but also expose hidden dataset biases, enhancing both performance and interpretability in robot learning.
FR-DETR achieves superior object detection accuracy in adverse weather conditions while cutting computational costs by refining features instead of images.
VLMs can identify distractors effectively, but they risk removing essential scene elements, revealing a significant flaw in current approaches to visual composition.
FastPano3D reconstructs detailed 3D indoor scenes from a single image in seconds, setting a new benchmark for speed and efficiency in 3D scene inference.
Achieving state-of-the-art RRSIS performance with just 2.4% parameter updates challenges the norm of full fine-tuning in multimodal models.
Unlocking the latent potential of timestep embeddings allows for a parameter-free, single-model approach to multi-task learning that rivals traditional, heavier methods.
Achieving state-of-the-art underwater image enhancement with a model that has only 4.23K parameters and processes images at over 600 FPS.
Latent visual reflection allows models to achieve superior fine-grained perception in long visual contexts while slashing inference time by nearly half.
HiRes achieves an impressive 85.8% identification accuracy from unconstrained images, setting a new standard for resistor value recognition in complex environments.
UniGP reveals that joint training of controllable generation and dense prediction can significantly enhance performance without the need for complex designs, outperforming specialized models.
Achieving accurate material decomposition in sparse-view DECT could revolutionize medical imaging by enabling safer, lower-radiation scans without sacrificing detail.
SkelEM achieves high-fidelity detail restoration in volume microscopy in under five steps, outperforming existing self-supervised methods.
CylindTrack achieves superior identity preservation in panoramic multi-object tracking by effectively modeling depth consistency and leveraging the unique topology of equirectangular images.
Achieving a Dice Similarity Coefficient of 79.4, LETT-NeXt redefines efficiency in 3D lesion segmentation without sacrificing accuracy.
Uncertainty estimates from pathology models can be reliably derived from ensemble disagreements, significantly enhancing trust in clinical applications.
CIPE-Dance is a game-changer, providing the largest dataset for dance video generation and enabling OmniDance to set new benchmarks in multimodal video synthesis.
Current T2I models fall short in scientific reasoning, but fine-tuning on the new SciIR dataset boosts performance by over 20%.
A self-supervised approach reveals that a shared canonical object frame can emerge from noisy video data, achieving pose estimation accuracy without manual annotations.
A learned covisibility module in Argus effectively eliminates global pose drift, setting a new benchmark for 3D reconstruction accuracy in indoor scenes.
Urban facade reconstruction can achieve superior geometric accuracy by integrating lightweight structural supervision, overcoming common pitfalls of traditional methods.
GeoEdit achieves unprecedented geometric accuracy and identity fidelity in object editing, overcoming the limitations of existing diffusion-based methods.
Aggregating gradients from multiple views can dramatically enhance 3D generation performance while cutting optimization steps in half.
LEIQ-Assessor not only sets a new standard for low-light image quality assessment but also reveals that multi-task learning can significantly enhance perceptual feature extraction.
RoamFlow achieves efficient image-goal navigation with reduced inference latency while outperforming traditional methods in trajectory generation.
Self-supervised learning can transform noisy LiDAR SLAM into a robust system by recursively refining local geometric representations.
A novel cross-spectral VTI system achieves superior accuracy and robustness in challenging environments by dynamically balancing visual and thermal data reliance.
Lightweight CNNs can achieve real-time steering predictions while reducing model size and complexity through automated hyperparameter optimization.
OP3DSG achieves state-of-the-art performance in 3D scene graph generation by capturing intricate spatial and functional relationships in real-world environments.
FalconTrack automates the generation of photorealistic labeled data, achieving 100% success in real-world tracking while traditional methods falter under pressure.
Achieving RGB-D performance with only monocular input, MyGO-Splat revolutionizes SLAM by integrating closed-loop geometric feedback for enhanced scale stability.
Achieving accurate 6-DoF pose estimation for fixed-wing UAVs without relying on CAD models could revolutionize UAV operations in complex environments.
Mega achieves a groundbreaking energy efficiency of 0.375 pJ/SOP, setting a new standard for convolutional spiking neural network accelerators.
Flow Splatting achieves superior image quality and faster rendering speeds by efficiently modeling dynamic scenes with 4D Gaussian representations.
High accuracy in VideoQA can be misleading, as many models perform better without visual input, revealing a dangerous reliance on textual shortcuts.
IDNet not only sets a new benchmark for IHD screening but also reveals that effective fusion of multimodal data can dramatically enhance diagnostic performance.
Integrating clinical risk factors into deep learning models can dramatically enhance the accuracy of coronary artery stenosis grading, outperforming traditional methods.
T2LDM++ generates realistic LiDAR scenes with rich geometric details, overcoming the limitations of existing models that struggle with insufficient training data and controllability.
A novel approach that combines attention mechanisms with variational image registration achieves superior performance while enhancing explainability and modularity in medical imaging.
Injecting intermediate text representations into the denoising process can dramatically enhance the alignment of text-to-image models with challenging prompts for unique objects.
Video diffusion models can revolutionize hand motion reconstruction, achieving unprecedented accuracy without relying on traditional detection or optimization techniques.
Forged watermarks in latent diffusion models are fundamentally limited by an irreducible distortion floor, revealing critical vulnerabilities in black-box settings.
LiDAR-based 3D object detectors can be compromised by targeting just a few critical spatial regions, revealing a significant structural vulnerability.
Achieving state-of-the-art 4D panoptic occupancy tracking, this method maintains temporal coherence in scene representations without the computational burden of recomputing volumetric data at each timestep.
Quadrotors can now navigate to specific objects in images by intelligently selecting viewpoints, significantly enhancing their operational capabilities in complex environments.
Sphere-VIO achieves unprecedented accuracy and efficiency in visual-inertial odometry for heterogeneous multi-camera systems, setting a new standard for real-time performance.
SyncCache accelerates audio-driven portrait animation by up to 4.12x while maintaining near-lossless visual quality, transforming the efficiency of DiT-based systems.
Achieving high-resolution photomosaics without training, PhotoQuilt outperforms traditional methods by balancing local detail and global coherence.
CogSENet achieves superior image deblurring by integrating semantic awareness and frequency fusion, outperforming traditional methods with fewer parameters.
Effective keyframe extraction can boost MLLM performance by over 2% on complex video-guided tasks, revealing a critical link between video understanding and procedural learning.
LightOnOCR-2-1B sets a new benchmark for Sinhala OCR, achieving a remarkable 1.05% CER while outperforming both open-source and commercial alternatives.
Despite integrating multiple detection modalities, the system fails to reliably reconstruct keystrokes in real-world scenarios, revealing critical limitations in current video surveillance technologies.
Bit-ViP achieves secure image obfuscation that preserves usability, making it a game-changer for privacy in computer vision applications.
High benchmark scores in video anomaly detection are misleading, with cross-dataset performance often reduced to chance levels, revealing a critical gap in deployable reliability.
Trust in deepfake detection systems can plummet as their competence wanes, revealing a critical link between performance and calibration that could redefine how we assess AI trustworthiness.
Ghosting artifacts can be significantly reduced by treating dynamic objects as persistent identities rather than fleeting appearances in monocular mapping.
Personalized predictions of driver behavior in dilemma zones can achieve over 93% accuracy by harnessing visual semantics and driver profiles.
MTD-Map achieves robust dynamic object removal and change detection in a single framework, cutting computational costs without sacrificing performance.
PL-LIT achieves state-of-the-art performance in thermal SLAM, even in challenging environments where traditional methods falter.
Session-level data leakage can inflate drone detection performance by over 6%, highlighting the critical need for rigorous validation practices in acoustic sensing research.
