Get improved room layout from image by using pixel meaning with SRW-Net

Better image classification in new domain by focusing on foreground with RobustViT

Train 3D segmentation network using simple bounding boxes with Box2Mask

Improve 3D scene reconstruction using monocular depth and normal cues with MonoSDF

Get 3D object from images captured under unknown conditions in the wild with SAMURAI

Adapt visual tasks to new domain without source domain data with DistillAdapt

Better object pose and tracking by propagating keypoints with CenterPoseTrack

Improve human motion estimate in video using physics with HUND+SO+GT+Dynamics

3D object pose from video scan without CAD model or training with OnePose

Explain image classifier by showing features in source and target domains with XSDA-Net

Enhance dark images in new untrained domain by simulating image pipeline for training

Classify small images accurately using little memory and CPU with ImageSig

Better object segmentation of unknown objects using region awareness with RAML

3D object reconstruction using a few views despite noisy camera poses with FvOR

Survey of learning from only a few examples with computer vision applications

Survey of filling in missing depth data using deep learning

Improve 3D object reconstruction using unsupervised learning on image collection with TARS

3D shape, segmentation, appearance, and object poses from images using neural fields with PNF

Better segment aerial photo by learning multi-resolution features with IFWM

Count crowds accurately even after add new domain with FLCB