Latest Articles for Computer Vision

A new method improves image classification by addressing misleading features.

A new method enhances the robustness of Mixture of Experts CNNs against adversarial inputs.

A novel method improves the ability of Vision-Language Models to adapt to new tasks.

A method for reconstructing maps from their branching curves in geometry.

SAM improves feature matching by considering group relationships in images.

New approach improves image quality and consistency in CcGANs using negative samples.

Examining the benefits of Diffusion Models for image classification and segmentation tasks.

Introducing a balanced approach to prune deep neural networks effectively.

A method for identifying changes in images taken from different viewpoints.

FastTCM-CR50 boosts scene text detection and spotting with limited data.

Strata-NeRF tackles multi-layered scene challenges for better virtual experiences.

A new method improves accuracy and efficiency in capturing 3D shapes of multiple people.

Spatial Transform Decoupling enhances accuracy in detecting tilted and rotated objects.

Research reveals how images influence answers in visual questions.

A new approach to improve machine learning models across various domains.

A new method to help AI learn continuously without losing past knowledge.

Exploring the role of saliency maps in improving segmentation accuracy.

New methods in image denoising enhance quality using fewer resources.

A new framework enhances hand pose estimation using multiple views.

New masking technique enhances video learning by utilizing motion for improved model training.

CoNe improves image classification by using similar images for training.

This article presents PMFS, a method for generating realistic images using autoencoders.

New techniques improve vanishing point estimation in images with unknown camera settings.

New methods improve identity recognition over time despite clothing changes.

A new method improves object position and orientation estimation in complex scenarios.

New methods improve accuracy for creating 3D models of challenging objects.

New methods improve image quality while speeding up the process.

A new metamaterial neural network design improves efficiency and speed in image processing.

New methods improve clarity and quality in hazy images.

This method improves the recognition of handwritten math symbols using semantic relationships.

A new approach to generate realistic materials for 3D graphics.

New models improve robotic grasping efficiency with fewer resources.

Exploring new methods for object recognition using self-supervised learning techniques.

A method improving mobile image processing through optimized quantization.

A study on the impact of long-term context in video action segmentation.

Equivariance helps neural networks recognize images consistently despite transformations.

This study explores how smaller receptive fields enhance action recognition in videos.

LOCATE identifies objects in images and videos without human input.

MEGA method improves video segmentation accuracy by integrating multiple data sources.

DiffSeg offers a new way to segment images without labeled data.