BEM offers a solution to improve models with imbalanced classes in semi-supervised learning.
― 7 min read
Computer Science - Computer Vision and Pattern Recognition
RSSA novel method improves the efficiency of creating human avatars.
― 6 min read
Examining biases in image generation and their societal impacts.
― 6 min read
TEAR efficiently aligns large 3D point sets, overcoming outliers and memory issues.
― 5 min read
A new training method enhances vision-language models' performance in zero-shot tasks.
― 7 min read
A new framework transforms image interpretation through open-vocabulary scene graphs.
― 7 min read
This article outlines a method for making digital twins of moving objects.
― 5 min read
A new dataset focuses on causal reasoning using 'Tom and Jerry' animations.
― 6 min read
Tiny drones enhance pest detection in farming, promoting sustainability and efficiency.
― 5 min read
LP++ improves vision-language model adaptation, especially in few-shot learning scenarios.
― 5 min read
SnAG improves video grounding accuracy and efficiency for longer videos.
― 5 min read
New methods improve safety by predicting out-of-sight pedestrian movements for autonomous vehicles.
― 6 min read
New datasets enhance the capabilities of Neural Architecture Search in real-world applications.
― 10 min read
Methods to improve rendering quality across varying scene sizes.
― 6 min read
A new method for easy and effective 3D avatar editing using a single image.
― 5 min read
A new framework enhances dynamic 3D content generation for animation and gaming.
― 5 min read
A new training method enhances the compositionality of vision-language models.
― 6 min read
A framework for aligning images of similar objects in 3D space.
― 7 min read
PriViLege framework enhances learning in Few-Shot Class Incremental Learning with large models.
― 6 min read
New method enhances camera movement control in text-to-video creation.
― 6 min read
A new approach for realistic traffic scenarios in autonomous vehicle testing.
― 6 min read
Surround-view cameras enhance driving safety but face challenges from optical artifacts.
― 6 min read
IISAN improves efficiency in multimodal recommendation systems while maintaining performance.
― 8 min read
A new method improves clarity in dark images for various applications.
― 5 min read
A new approach to enhance learning when labeled data is scarce.
― 5 min read
Examining the reliability of visual explanations in computer vision models.
― 5 min read
Bi-LORA improves detection of AI-generated images using vision-language models.
― 7 min read
ASTRA model improves accuracy in identifying actions during soccer matches.
― 6 min read
A new approach improves efficiency in detecting objects using event-based cameras.
― 5 min read
New method improves object detection using fewer labeled data.
― 6 min read
A new method improves imaging of sensitive samples in low light.
― 6 min read
MotionChain creates realistic human movements through conversation, combining text, images, and motion data.
― 6 min read
A new method enhances low-resolution images for better recognition tasks.
― 7 min read
Introducing EARL: a new method for effective online continual learning.
― 6 min read
A new approach estimates 3D poses from multiple views without ground-truth data.
― 7 min read
BodyMAP combines depth and pressure data to enhance care for bedridden patients.
― 7 min read
A new method allows artists to customize generated images with greater control.
― 7 min read
A novel method for improving multi-label classification using synthetic image generation.
― 6 min read
A new method enhances object detection in autonomous vehicles using combined sensor data.
― 8 min read
LidarDM generates realistic LiDAR videos for autonomous vehicle training and testing.
― 5 min read