Reconstructing resolution-independent materials represented by procedural graphs from a coarse 3D model and a photo of an interior

3D reconstruction of garments from single in-the-wild photos

Learning to model secondary motion, e.g., clothes for a danching human, via equivariant transformations

3D reconstruction with deformable animations from casual video collections

Generating 3D avatars from VR headset cameras

A new model for camera noise based on normalizing flows

Generating images from object detection layouts via VQ-VAE and a Transformer

3DMM as a geometry prior for implicit function representations of facial shape and texture

Automated tongue animations that match recorded speech

Monocular reconstruction in PIFu style with disentangled colors

Controllable 3D morphable face model with a coarse-to-fine structure

3D morphable face model based on neural implicit fields

Virtual try-on (garment transfer) in varied poses from a new dataset of dancing videos

GAN for generating photorealistic humans with controllable clothing

Improved pretraining for vision transformers based on formula-generated synthetic fractal images

New approach to training on a mixture of synthetic and real data for crowd counting

New synthetic dataset used to train a state of the art model for eyeglass removal

Generating synthetic samples to improve few-shot learning

Simulated+Unsupervised (S+U) learning is proposed to learn a model to improve the realism of a simulator’s output using unlabeled real data, while preserving the annotation information from the simulator

3D reconstruction of hand-held objects from a single RGB image without knowing their 3D templates.

An extensible library for generating 3D synthetic data with Blender that just works.

Adaptive Batch Normalization (AdaBN) to increase the generalization ability of a DNN for domain adaptation tasks.

A novel unsupervised domain adaptation framework based on domain-specific batch normalization in deep neural networks

A new representation learning approach for domain adaptation, in which data at training and test time come from similar but different distributions

Method to learn a nonlinear transformation that aligns correlations of layer activations in DNN for a target domain that is unlabeled and requires unsupervised learning

A discriminatively-trained Cycle-Consistent Adversarial Domain Adaptation model that adapts representations at both the pixel-level and feature-level, enforces cycle-consistency while leveraging a task loss, and does not require aligned pairs

Image-to-image translation, where each patch in the output reflects the content of the corresponding patch in the input, independent of domain

A two-stream acrchitecture to expliclitly model the shift in features from one domain to the other

A data generation pipeline for creating semi-realistic synthetic multi-object videos for ML tasks using PyBullet and Blender.

StyleGAN2-ADA: adaptive discriminator augmentation that helps use StyleGAN2 with small datasets

The original StyleGAN from NVIDIA allowing to mix and match facial features at multiple levels

The original progressively growing GAN from NVIDIA

A common adversarial loss function based on least squares

SAGAN: GANs with self-attention feature maps

A common adversarial loss function based on Wasserstein's distance

The original paper on GANs

StyleGAN3: making StyleGAN translation and rotation invariant

StyleGAN2: fixing several shortcomings of StyleGAN and taking it one step further

InfoGAN: disentangling representations by optimizing a mutual information objective

Classical GAN for superresolution

First successful convolutional GAN for generating images

Large-scale GANs for generating varied class-conditional images, with hundreds of classes

pix2pix: first GAN-based style transfer model for paired data

CycleGAN: style transfer without paired data via cycle consistency

AdaIN: a simple and very powerful idea for neural style transfer

A new weight normalization technique that generally improves GAN training

Bidirectional GANs that learn to generate new samples and extract features at the same time

Superresolution GAN improved in both architecture and loss functions

pix2pixHD: high-resolution style transfer for paired datasets

FUNIT: style transfer with styles defined by a few sample images

MUNIT: disentangling style and content for style transfer

Real-SR improved with better discriminators and synthetically generated data

Learning degradation kernels for superresolution with kernel prediction networks

ZSSR: learning downscaling kernels for superresolution from a single image

KernelGAN: adversarial training of downscaling kernels for superresolution

Head pose estimation model trained on synthetic data with syn-to-real domain adaptation

A new benchmark for synthetic-to-real domain adaptation

Joint alignment of synthetic and real domains for robotics

A new technique to bridge source and target domains in adversarial domain adaptation

A meta-learning scheme for unsupervised domain adaptation

Unsupervised domain adaptation with feature decomposition and alignment

Object tracking with strong domain generalization capabilities

Classification and survey of visual domain adaptation techniques, >200 references

Survey on domain adaptation for LiDARs with ~80 references

Domain adaptation with Siamese networks

Domain adaptation based on optimal transport

A domain adaptation model with shared and private encoders and a shared decoder

Unsupervised domain adaptation with gradient reversal

A survey of domain adaptation with ~80 references

In this paper, we propose a new assumption, generalized label shift (GLS), to improve robustness against mismatched label distributions