CVPR 2020 CutBlur and MoA: A powerful data augmentation for various low-level vision tasks. Paper: arxiv.org/abs/2004.00448 Github: github.com/clovaai/cutblur Slides: bit.ly/3fluybm Youtube: youtu.be/MQR5mnsCvbM

Our two papers were accepted at #icml2020!! Congrats Seong Joon Oh, SanghyukChun , YoungjungUH , Jaejun Yoo , Yunjey Choi and #Clova_AI_Research members.

Happy to announce our CVPR 2020 paper on a new evaluation for weakly-supervised object localization (WSOL)! WSOL methods have many issues. E.g. often not truly "weakly-supervised". We fix the issues. Authors: Junsuk Choe Seong Joon Oh Seungho Lee Zeynep Akata Hyunjung Shim (1/13)

Training Generative Adversarial Networks with Limited Data pdf: research.nvidia.com/sites/default/… project page: research.nvidia.com/publication/20… HT 𝔊𝔴𝔢𝔯𝔫

Glad to share our new work "Rethinking the Truly Unsupervised Image-to-Image Translation". We tackle image-to-image translation **without any human supervision**. For details, please check kyungjune's original thread!

Happy to share that our new #optimizer, #adamp, was released with official pytorch code and #pip install. AdamP tackles the excessive increase of weight norm by using #momentum with #normalization. SanghyukChun Seong Joon Oh YoungjungUH clovaai.github.io/AdamP/

Moving away from negative pairs in self-supervised representation learning: our new SotA method, Bootstrap Your Own Latent (BYOL), narrows the gap between self-supervised & supervised methods simply by predicting previous versions of itself. See here: bit.ly/30MCugQ

New video 🥳 about unsupervised image-to-image translation. This model removes the need for any and all labels by jointly training a self-clustering guiding network along the translation model! Watch now 🔥 youtu.be/sEG8hD64c_Q @KyungjuneB Yunjey Choi YoungjungUH Jaejun Yoo

The official Pytorch code of our #Rebias was released, presented at #icml2020! Rebias addresses the bias issue in real-world data (e.g. predicting a snowmobile with background snow) using a bias-designed network. github.com/clovaai/rebias SanghyukChun Seong Joon Oh

Real-world datasets are biased; ML models will be biased w/ biased data. Our paper on bias generalization, ReBias, will appear at #icml2020 Paper arxiv.org/abs/1910.02806 Code github.com/clovaai/rebias Talk youtu.be/Z_e_XWJWeJ8?t=… w/ Hyojin Bahng Sangdoo Yun, Jaegul Choo Seong Joon Oh

PolyFit: Perception-aligned Vectorization of Raster Clip-art via Intermediate Polygonal Fitting pdf: cs.ubc.ca/labs/imager/tr… project page: cs.ubc.ca/labs/imager/tr…

We will be presenting our work on Reliable Evaluation of Generative models this Friday 17, July at 12am and 11am CEST at #ICML2020. Recorded presentation at: youtube.com/watch?v=_XwsGk… Drop by the Q&A session if you are interested. Seong Joon Oh YoungjungUH Yunjey Choi Jaejun Yoo

1. Reliable fidelity and diversity metrics for generative models Ferjad Naeem , Seong Joon Oh , YoungjungUH , Yunjey Choi , Jaejun Yoo - Paper: arxiv.org/abs/2002.09797 - Github: github.com/clovaai/genera… - Youtube: youtu.be/_XwsGkryVpk - ICML interactive: icml.cc/virtual/2020/p…

🔥New Video🔥 LambdaNetworks capture long-range interactions as linear functionals🤯 Super complicated, basically Transformers without the giant memory requirements 🥳 New SotA on ImageNet! 💪 Watch Now! youtu.be/3qxJ2WD8p4w #ICLR2021

Happy to announce our new Pooling-based Vision Transformer (PiT) is just released in ArXiv and GitHub (including pre-trained weights!) Rethinking Spatial Dimensions of Vision Transformers ArXiv: arxiv.org/abs/2103.16302 GitHub: github.com/naver-ai/pit See the thread for details

StyleMapGAN: Exploiting Spatial Dimensions of Latent in GAN for Real-time Image Editing pdf: arxiv.org/pdf/2104.14754… abs: arxiv.org/abs/2104.14754 github: github.com/naver-ai/Style…

Happy to share the code and paper of #StyleMapGAN of #NAVER_AI_Lab was released. Please check it out! github: github.com/naver-ai/Style… arxiv: arxiv.org/abs/2104.14754 youtube: youtu.be/qCapNyRA_Ng Congrats to Hyunsu Kim, Yunjey Choi , Junho Kim, Sungjoo Yoo, YoungjungUH

Glad to share our updated AFHQ dataset. We rebuild the original dataset using high-quality resize filtering (i.e., Lanczos resampling). We thank Alias-Free GAN authors for their suggestion and contribution. The update dataset can be found at github.com/clovaai/starga….

Yes. Generator knows what D should learn. The feature map from G is as useful as GT segmentation map. This work will be in #ECCV2022 (European Conference on Computer Vision #ECCV2026 ). Since #ReXNet, this is the best name that I gave :). github.com/naver-ai/GGDR @ClovaAiLab Yunjey Choi

3D-aware Blending with Generative NeRFs abs: arxiv.org/abs/2302.06608 project page: blandocs.github.io/blendnerf