Latest Articles for Computer Vision

New benchmark evaluates object-centric learning against image distortions.

Exploring effective methods for realistic shape reconstruction using 3D Gaussians.

This article discusses enhancing robot tracking of shiny objects using depth measurement improvements.

Learn how domain adaptation improves point cloud segmentation performance.

Examining the risks associated with leading computer vision models and their effectiveness.

New methods improve human pose and shape estimation from 3D data.

A new method enhances object detection by addressing class imbalance in datasets.

Explore the latest methods in producing high-quality images from different viewpoints.

New techniques enhance face parsing despite occlusions.

GEST bridges the gap between visual elements and language for clearer content creation.

A method to enhance target recognition from SAR images by reducing background interference.

R-LPIPS improves image similarity assessment against adversarial examples.

This study explores SSL models for classifying underwater sonar images.

A method enhancing semi-supervised learning by distinguishing known and unknown data.

A new method enhances referring image segmentation using weakly supervised learning.

This study introduces constrained margins for better generalization predictions in DNNs.

A four-step method for adapting object detectors to new environments.

A new model improves accuracy in estimating human body poses from video data.

Near-online methods enhance object tracking in video analysis.

A new method creates a dataset for predicting 3D facial points from 2D images.

A method to refine text recognition from single handwritten images.

A new framework enhances anomaly detection in visual data without specific training.

A new approach enhances accuracy in answering questions about text in images.

Enhancing vision transformers to handle diverse and out-of-distribution data.

Researchers improve CNN and Transformer models to resist adversarial examples.

A new method uses AI for real-time environment mapping and updates.

A new approach enhances crowd analysis methods for better detection and counting.

Assessing methods for finding similar fashion items in online shopping.

New methods improve image generation from limited examples, enhancing quality and diversity.

A new approach enhances blob detection accuracy in noisy images.

PROOFREAD enhances visual question answering using knowledge from large language models.

New datasets and algorithms improve tracking of underwater objects.

This article discusses RSCDT's role in improving signed image classification.

A novel approach turns facial photos into human-like drawings using advanced techniques.

A new method for better evaluating object proposals in vision and language tasks.

D-LKA Attention improves the accuracy and efficiency of medical image segmentation.

A new model improves medical image segmentation by focusing on high-frequency details.

A novel generator improves image creation across various categories using optimal transport methods.

A novel method using generative models to enhance image segmentation tasks.

New methods improve player and ball tracking in soccer matches.