BEiT-large (pretrained on ImageNet-22k) achieves SoTA on ADE20K (57.0 mIoU), and 88.6% top-1 accuracy on ImageNet-1k. Self-supervised learning is the future. Code and pretrained checkpoints: github.com/microsoft/unil… twitter.com/ak92501/status…

○ TensorRT is an SDK for high-performance deep learning inference and with TensorRT 8.0, you can import models trained using Quantization Aware Training (QAT) to run inference in INT8 precision…

Epistemic Neural Networks pdf: arxiv.org/pdf/2107.08924… abs: arxiv.org/abs/2107.08924 introduce the epistemic neural network (ENN) as an interface for uncertainty modeling in deep learning

QuantumNAS: Noise-Adaptive Search for Robust Quantum Circuits pdf: arxiv.org/pdf/2107.10845… abs: arxiv.org/abs/2107.10845 For QML tasks, QuantumNAS is the first to demonstrate over 95% 2-class, 85% 4-class, and 32% 10-class classification accuracy on real…

Made a quick Colab notebook to explore OpenAI's CLIP and visualize which parts of an image activate given a certain caption. colab.research.google.com/github/kevinza…

Towards Generative Video Compression pdf: arxiv.org/pdf/2107.12038… abs: arxiv.org/abs/2107.12038 a neural video compression method based on GANs that outperforms previous neural video compression methods and is comparable to HEVC in a user study

A Realistic Simulation Framework for Learning with Label Noise pdf: arxiv.org/pdf/2107.11413… abs: arxiv.org/abs/2107.11413 github: github.com/deepmind/deepm… a simulation framework for generating realistic instance-dependent noisy labels via a pseudolabeling…

This is an official implementation for "Contextual Transformer Networks for Visual Recognition". - GitHub - JDAI-CV/CoTNet: This is an official implementation for "Contex...

Go Wider Instead of Deeper pdf: arxiv.org/pdf/2107.11817… abs: arxiv.org/abs/2107.11817 best model outperforms Vision Transformer (ViT) by 1.46% with 0.72× trainable parameters

Official implementation of the CVPR 2022 paper "DETReg: Unsupervised Pretraining with Region Priors for Object Detection". - GitHub - amirbar/DETReg: Official implementation of th...

We’ve just added built-in citation support to GitHub so researchers and scientists can more easily receive acknowledgments for their contributions to software. Just push a CITATION.cff file and we’ll add a handy widget to the repo sidebar for you. Enjoy! 🎉

We’re releasing Triton 1.0, an open-source Python-like programming language for writing efficient GPU code. OpenAI researchers with no GPU programming experience have used Triton to produce kernels that are 2x faster than their PyTorch equivalents. opena…

Today we are releasing the Open Buildings Dataset, a new open-source dataset containing the locations and footprints of >500M buildings with coverage across Africa, which can support numerous scientific and humanitarian applications. Read more at goo.gle/3751HVR