Latest Articles for Computer Vision

A new method improves action recognition from skeleton data using advanced pooling techniques.

This article discusses using image captions to find videos efficiently.

New model improves depth estimation using event camera data through efficient algorithms.

This study combines RGB-D cameras and IMUs for better motion estimation.

New method enhances how machines navigate and understand language commands.

New model improves vehicle environment recognition using cameras and LiDAR.

Introducing the ViOCRVQA dataset for improved visual question answering in Vietnamese.

ShapeMoiré improves image quality by effectively removing unwanted moiré patterns.

Llip enhances how images are matched with diverse textual descriptions.

A concise look at hallucinations in MLLMs and strategies to improve reliability.

SGD-PH combines first-order and second-order methods for better model training performance.

A comprehensive dataset of street view images for geolocation projects worldwide.

A model adapts to various image tasks using minimal examples.

New method enhances shadow removal in images through deep learning and transformers.

New methods enhance visual scene analysis using efficient coding techniques.

Study reveals insights on the balance between visual and textual inputs in VLMs.

MV-RGBT offers a realistic dataset for evaluating RGBT tracking methods.

This article explores medial parametrization, a technique for describing complex flat shapes.

New techniques reduce memory access and boost performance in deep learning models.

Introducing LVOS: a dataset for tracking objects in long videos.

Kite improves transferability estimation for better model selection in transfer learning.

A new approach enhances multi-subject image generation using layout manipulation.

A new method improves object recognition by encouraging compositionality in image representations.

Wake Vision enhances person detection for TinyML with a vast dataset.

Explore the rise and efficiency of Vision Transformers in image processing.

M3Net enhances LiDAR segmentation for self-driving cars by integrating diverse datasets and sensors.

New dataset improves model performance on multi-image tasks.

Differentiable Particles approach revolutionizes how robots handle changing shapes.

A new method creates complex 3D scenes from straightforward videos with multiple objects.

A new method enhances vision-language models without complex training.

Idefics2 showcases improvements in vision-language processing through innovative design choices.

Exploring the connection between deep generative models and the manifold hypothesis.

A new method enhances image descriptions for training AI models.

A new approach tackles action segmentation in lengthy videos using optimal transport.

UnSAMFlow improves optical flow estimation using segment-level information for better accuracy.

Discover how CPEA method enhances image classification with minimal data.

A new approach improves AI's ability to learn from limited examples.

A new method enhances accuracy in estimating human poses from 2D images.

Enhancing diffusion models by adding LoRA to attention layers for better images.

A new method for quick camera exposure adjustments using deep reinforcement learning.