Paper Library

[hyoseok1223]

Paper List

Paper	Topic	Links	Conference
Swin Transformer: Hierarchical Vision Transformer using Shifted Windows	Architecture	arxiv	ICCV '21 (Best Paper)
MobileViT: Light-weight, General-purpose, and Mobile-friendly Vision Transformer	Architecture	arxiv	ICLR '22
Masked Autoencoders Are Scalable Vision Learners	Architecture	arxiv	CVPR '22
MAXIM: Multi-Axis MLP for Image Processing	Architecture	arxiv	CVPR '22 (Oral)
A ConvNet for the 2020s (ConvNeXt)	Architecture	arxiv	CVPR '22
DiffStride: Learning Strides in Convolutional Neural Networks	CNN	arxiv	ICLR '22 (Oral, Outstanding Paper)
GAN-Supervised Dense Visual Alignment (GANgealing)	GAN-Supervised Learning	arxiv	CVPR '22 (Oral, Best Paper Finalist)
ReStyle: A Residual-Based StyleGAN Encoder via Iterative Refinement	GAN Inversion	arxiv	ICCV '21
Real Image Inversion via Segments	GAN Inversion	arxiv	ECCV '22
3D face reconstruction with dense landmarks	3D Face Reconstruction	arxiv	ECCV '22
MegaPortraits: One-shot Megapixel Neural Head Avatars	Nerural Talking Head	arxiv	ACMM '22
Diffusion Models Beat GANs on Image Synthesis	Image Synthesis	arxiv	NIPS '21 (Spotlight)
Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domains	Positional Encoding	arxiv	NIPS '20 (Spotlight)
ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness	Texture Bias	arxiv	ICLR '19 (Oral)
The Origins and Prevalence of Texture Bias in Convolutional Neural Networks	Texture Bias	arxiv	NIPS '20
Revisiting the Calibration of Modern Neural Networks	Calibration	arxiv	NIPS '21
Exploring Simple Siamese Representation Learning (SimSiam)	Representation Learning	arxiv	CVPR '21 (Best Paper Honorable Mentions)
Categorical Reparameterization with Gumbel-Softmax	Reparameterization Trick	arxiv	ICLR '17
Revisiting ResNets: Improved Training and Scaling Strategies	Training Strategy	arxiv	NIPS '21 (Spotlight)
Meta Pseudo Labels	Semi-supervised Learning	arxiv	CVPR '21
Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time	Weight Averaging	arxiv	ICML '22