Latest Articles for Computer Vision

A method to generate questions from images and captions for better AI interaction.

New methods improve accuracy and consistency in image recognition models.

A novel approach using instance-wise data augmentation for better adversarial model robustness.

A new method enhances Vision Transformers for better image understanding with fewer resources.

A structured approach for effective sensor positioning in robotic vision tasks.

A new method enhances image realism through 3D shape control in diffusion models.

AVIS system improves visual question answering through structured workflows and transition graphs.

Exploring the potential of event cameras in enhancing pedestrian detection for autonomous vehicles.

A benchmark for assessing image similarity based on user-defined conditions.

New method improves depth estimation using dual-pixel sensors in various imaging devices.

A method to create realistic 3D shapes using only 2D data.

A new method enhances text removal techniques in images.

Adversarial examples can confuse object detection systems, revealing security gaps.

A new model enhances action detection speed and accuracy in real-time video analysis.

Anisotropy affects the performance of Transformer models across various data types.

A new method enhances how models grasp image-text relationships.

OCAtari focuses on game objects for better machine learning.

A new method employs neural architecture search to improve face forgery detection.

A new model improves the link between images and their text descriptions.

A new method creates lifelike 3D avatars from just one photo.

New methods enhance quality and speed in text-to-image models.

This study explores how AI can learn words by connecting them to images.

A new method predicts 3D shapes from single RGB images using depth data.

TomoSAM streamlines 3D image segmentation, enhancing efficiency and accuracy for researchers.

Research enhances sketch recognition for improved 3D shape matching.

P2D improves 3D object detection in self-driving cars using motion prediction.

New methods improve image quality using real-world light field data.

Ground-VIO improves vehicle pose estimation using camera-ground relationships.

A new model enhances how machines recognize images by blending global and local features.

A new strategy ensures equal representation of data types in machine learning.

A new approach to enhance trust in object detection through reliable calibration techniques.

A new method reveals how eye reflections can reconstruct 3D environments.

MaskDiT enhances diffusion model training efficiency while maintaining image quality.

A study on Visual Foundation Models' performance under real-world distortions in segmentation tasks.

DiffAug enhances image recognition systems through innovative noise techniques.

Introducing CANN, a method for accurate visual localization using local features.

A new method enhances image generation from text by properly linking entities and modifiers.

New methods enhance segmentation of surgical instruments for improved robotic surgeries.

A new method enhances image analysis for biomedical applications.

FETNet improves scene text removal methods for better privacy and image restoration.