Latest Articles for Computer Vision

A flexible approach to improve image quality without prior knowledge of issues.

A new method enhances image retrieval using Soft Orthogonality and Vision Transformers.

This article examines vulnerabilities in semantic segmentation models and proposes solutions.

An overview of MLLMs and their applications in AI.

A new method for identifying OOD data in machine learning models.

Evaluating how environmental changes affect image recognition model performance.

ProRes offers a flexible solution for diverse image restoration challenges.

A look at the impact of manifold contrastive learning on machine learning models.

New method blends human shapes to predict movements accurately.

New methods improve defect detection using limited data in industrial products.

Exploring the benefits and challenges of semi-supervised object detection methods.

A new method for generating clear labeling instructions for image datasets.

A new method improves detection of known and unknown objects.

A new approach improves low-light image clarity without needing paired data.

FLOWER addresses few-shot learning and catastrophic forgetting in machine learning models.

This study enhances smaller AI models using knowledge from larger language models.

New method uses image mining for training without annotations.

Improving pedestrian detection models to retain knowledge while learning new datasets.

A new approach enhances template matching efficiency in computer vision tasks.

New techniques improve view generation for large, real-world scenes.

A new method improves control over image generation while maintaining efficiency.

A look into how zero-shot learning classifies new data without prior examples.

An analysis of ImageNet versus LAIONet and their effects on model accuracy.

Introducing xAI-CycleGAN for faster, high-quality image style transformations.

A new method enhances scene understanding for autonomous vehicles using instance queries.

A new method enhances 3D modeling in low-texture scenes without keypoint detection.

New dataset and metrics improve assessment of albedo quality in images.

New deep mesh deformation method enhances abdominal organ segmentation accuracy.

A new method enhances fine-grained action recognition using skeleton data.

Shikra enables natural conversations about specific areas in images.

A new method enhances speed and accuracy in segmenting unseen objects for robots.

New approach improves machine learning with semantic positive pairs.

A new method reduces label noise by focusing on positive and unlabeled data.

This study evaluates how well video retrieval models understand captions and video content.

New methods for accurate real-time environment mapping in robotics and augmented reality.

Examines the impact of foreground-background imbalance on object detection performance.

A new method improves how AI understands images and text.

A new method enhances how models learn from data transformations.

A new approach to improving model performance in imbalanced datasets.

A study on hand-crafted and deep learning methods for 3D object recognition.