Latest Articles for Computer Vision

This study analyzes the performance of KANs in continual learning using the MNIST dataset.

A new method to improve deep learning models against adversarial attacks.

FullAnno enhances image annotations for better multimodal model training.

A new method for video denoising that improves quality without paired datasets.

New framework enhances AI planning by improving visual understanding and reasoning.

Combining textual and visual data improves few-shot learning performance.

Study reveals context bias impacts object detection performance across different environments.

A new method enhances text reading accuracy from unclear images.

Exploring a new dataset for non-rigid point cloud registration.

New depth estimation methods improve robot interactions with complex environments.

A new approach enhances video question answering through scene text recognition.

This article discusses the effectiveness of Structure from Motion for accurate 3D modeling.

A new method enhances liver vessel segmentation in medical imaging.

This article examines methods to improve UDA performance in image classification tasks.

A new method improves accuracy in segmenting RGB-D images with depth errors.

New method improves depth map accuracy using multiple viewpoints.

Adaptive attention techniques boost performance and reduce resource demands in LVLMs.

Research improves data generation in machine learning using synthetic methods for clearer explanations.

A new method improves camera synchronization through innovative use of trifocal tensors.

This article investigates how VLMs perceive color, shape, and meaning in images.

A new method enhances accuracy in underwater image classification by isolating key features.

New methods enhance video summarization accuracy while reducing computational costs.

Examining strategies for improving feature learning in imbalanced datasets.

A new framework improves camera pose estimation in various environments.

Exploring the balance between adversarial threats and proactive measures in machine learning.

A look at using smaller adjustments for large pre-trained models.

Study enhances accuracy in hand gesture recognition using ultrasound data.

A method for accurate camera calibration using a single spherical mirror.

SMART enhances open-vocabulary segmentation by improving mask classification techniques.

Exploring how humans and machines perceive faces in random patterns.

Combining graph neural networks and variational autoencoders enhances image classification accuracy.

New method enhances object detection for unknown items and relationships.

A new benchmark improves evaluations of models counting objects using language prompts.

A new method combining models to improve unsupervised domain adaptation in segmentation tasks.

This study highlights the importance of object detection in construction zones for self-driving cars.

DALNet improves image segmentation accuracy using both visual and textual features.

LaPose improves object positioning using standard RGB images, addressing key challenges.

New models enhance CNN performance against corrupted images using human visual processing methods.

Innovative methods for improving image accuracy and clarity through quaternion tensor techniques.

SGDrop helps CNNs learn better from limited data by broadening their focus.