Latest Articles for Computer Vision

New method simplifies video generation using existing image models without extensive training.

A new method enhances video forecasting by separating scene elements for better predictions.

RagLLaVA enhances multimodal models, improving accuracy in complex data tasks.

A new framework for optimizing complex geodesic convex problems in various fields.

SuperVINS enhances robot navigation with deep learning techniques for improved mapping.

Robots improve task performance by learning from novel experiences and intrinsic rewards.

A new method improves how AI models interpret spatial and temporal relationships.

MAFT+ framework enhances object segmentation using collaborative optimization of vision and text.

Joint Neural Networks tackle challenges in recognizing objects from minimal examples.

This article examines the relationship between model size and performance in multimodal language models.

A new method improves 3D object detection using LiDAR data without labels.

A project using spiking neural networks for recognizing ASL gestures.

A new framework aims to reduce hallucinations in LVLMs through active retrieval.

A framework to reduce false outputs in language-vision models across multiple languages.

RPrDepth uses single images for accurate depth estimation leveraging rich-resource data.

A new method enhances text detection and recognition in challenging conditions.

A new method to enhance data variety for improved model performance.

Discover how Local Binary Patterns enhance image texture analysis.

This model predicts object movement and analyzes video content effectively.

Improving accuracy in estimating head orientations for various applications.

Integrating camera properties improves self-supervised depth estimation accuracy.

A new approach enhancing machine understanding of visual data from diverse sources.

A new approach to analyzing how image models withstand input changes.

New methods improve accuracy in detecting salient objects in high-resolution images.

A look at IG-SLAM and its impact on real-time mapping technology.

A new framework improves object recognition in images using text.

A new method enhances image restoration by fine-tuning models efficiently.

New method enhances privacy for vision transformers in machine learning.

FBINeRF enhances 3D rendering for regular and fisheye cameras.

A new method enhances semi-supervised learning using OOD data effectively.

A new framework improves how we assess image captions using language models.

Introducing a model to clarify ambiguities in binary edge images.

Examining vulnerabilities in vision transformers and downstream models through transfer attacks.

A new method improves tracking accuracy in challenging 3D environments.

CAFormer enhances object tracking by merging visible light and thermal infrared images.

This method enhances visual reasoning by implementing verification at each reasoning step.

A method to estimate 3D poses of closely interacting individuals using avatars.

A new method improves the prediction of human movement through past data analysis.

Exploring new methods to improve neural surface reconstruction using diverse features.

A new model for realistic face swapping using advanced techniques.