Latest Articles for Computer Vision

A new method identifies attacks on deep hashing systems without labeled data.

New techniques help AI agents predict future scenes for better navigation.

Two new methods improve normal estimation from point clouds using unsupervised techniques.

A new metric analyzes bias in image captioning for fairer AI systems.

POMP enhances visual recognition by using language-based prompts for improved image classification.

A new method enhances object tracking and segmentation in video analysis.

Study reveals how control policies affect vision model performance in robotics.

DetCLIPv2 enhances object recognition using image-text pairs for better flexibility.

A new method improves image demosaicing performance with smaller models.

Introducing HDT-HDR: A new method to reduce ghosting in HDR images.

This method improves adversarial attacks by altering features without changing visual quality.

A new benchmark for recognizing small and dense text in videos.

Investigating the impact of model size on remote sensing tasks.

StageInteractor improves object detection through innovative label assignment and resource optimization.

Proposed variations of Squeeze and Excitation modules boost CNN functionality and accuracy.

New RNN cell designs improve image classification performance significantly.

A new method enhances text editing in images using a dual encoder approach.

VAL-PAT improves recognition and search for pedestrians using images and text.

This study highlights methods to enhance image classifier performance through zooming and addressing biases.

A new model simplifies training and improves neural network performance across various tasks.

A new method improves training for identifying camouflaged objects with diverse images.

A new method enhances image segmentation by optimizing resizing factors efficiently.

MOST method locates multiple objects in images without extensive labels.

A new method improves point cloud registration efficiency with compressed SDFs.

A new method predicts how liquid drops behave on surfaces.

A method using stereo cameras for accurate vehicle speed estimation.

SportsMOT offers valuable data for tracking athletes in basketball, volleyball, and football.

This article presents a quicker method for optimizing camera positions in 3D modeling.

A novel method boosts object detection with limited data training.

CrowdSim2 offers a synthetic dataset to improve object detection models.

A new approach in fashion models focuses on fine-grained attributes.

MUREN improves human-object interaction detection through better context sharing.

Exploring the potential of Few-Shot Learning in semantic segmentation.

A new method for accurately locating corneal reflections in eye images.

A new algorithm enhances efficiency and accuracy in annotating 3D point clouds.

Introducing a straightforward approach for effective keypoint detection in computer vision.

A new method enhances face inpainting, preserving identity and quality.

Resource-efficient CLIP reduces computing needs while maintaining strong performance in image-text tasks.

Learn how data augmentation improves machine learning performance with imbalanced data.

NoisyTwins enhances StyleGANs for better image generation from limited datasets.