Github's hot AI project list for July has a sequel! Nuwa Unlimited Edition can only be ranked third?

The second edition of July’s most popular AI research list is here!

The previous list was sorted based on Twitter likes, retweets and Github stars. However, some netizens complained that the number of likes on Twitter could be brushed by machine, and users who liked Twitter could not You must know research.

This time, the author @bycloudai learned from experience. This time the indicator does not need to be the number of likes on Twitter, but the number of stars on Github.

Although the indicators this time are not very professional, compared to Twitter, you can browse Github The number can be considered to be more closely related to AI research.

In addition, the author also made it clear at the beginning of the list that this list is a wild list that he made for fun and is for entertainment only.

Having said that, this "Sequel List" basically maintains the style of the previous issue. The top 10 are given the resource library address, paper link, title, author and publishing unit.

Let’s take a look at what research is on the list of the most popular AI research in July in this "improved version"~

Top1 ：YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors

Author: Chien-Yao Wang, Alexey Bochkovskiy, Hong -Yuan Mark Liao Organization: "Academia Sinica" Institute of Information Science Github stars: 3.8k

Summary: YOLOv7 outperforms all known object detectors in speed and accuracy from 5FPS to 160FPS, and has the highest accuracy of all real-time object detectors above 30FPS on GPU V100 at 56.8% AP . The YOLOv7-E6 object detector (56 FPS V100, 55.9% AP) is 509% faster and more accurate than the Transformer-based detector SWIN-L Cascade-Mask R-CNN (9.2 FPS A100, 53.9% AP). 2%.

It is worth mentioning that the first author of this paper, Chien-Yao wang (Wang Jianyao), is a graduate student in machine engineering at Columbia University and is now a Chrome developer at Intel software engineer.

Top2：Text-Guided Synthesis of Artistic Images with Retrieval-Augmented Diffusion Models

Authors: Robin Rombach, Andreas Blattmann, and Bjorn Ommer Institution: Ludwig-Maximilians-Universität München

Github Stars: 2.4k

Abstract: New architecture has recently improved generative image synthesis, enabling excellent visual quality in a variety of tasks. Of particular note is the field of “AI-Art”. By combining speech and image synthesis models, so-called "cue engineering" was established, in which carefully selected and combined sentences are used to achieve a certain visual style in the synthesized images.

This paper proposes an alternative approach based on the Retrieval Enhanced Diffusion Model (RDM). In RDM, a set of nearest neighbors is retrieved from an external database during training for each training instance, and the diffusion model is conditioned on these informative samples.

Let’s take a look at the effect of the AI-Art works in the paper~

Top3：NUWA-Infinity: Autoregressive over AutoregressiveGeneration for Infinite Visual Synthesis

## Authors: Wu Chenfei, Liang Jian, Xiaowei Hu and other institutions ：Microsoft Research Asia, Peking University, Microsoft Azure AIGithub Stars: 2.4k

Ranked third is NUWA INFINITY, the AI ​​masterpiece of the all-China class that has attracted much attention before.

##4. Training Transformers Together (1K stars)

Authors: Alexander Borzunov, Max Ryabinin, Tim Dettmers and other institutions: Russian State Higher School of Economics, University of Washington, etc.

5. Theseus: A Library for Differentiable Nonlinear Optimization (791 stars)

Author: Luis Pineda, Taosha Fan, Maurizio Monge organization: Meta AI, Reality Labs Research

6, k-means Mask Transformer (704 stars)

Issuing unit: Johns Hopkins University, Google Research Resources Library: https://github.com/google-research/deeplab2 Paper: https://arxiv.org/abs/2207.04044v1

7、XMem: Long- Term Video Object Segmentation with an Atkinson-Shiffrin Memory Model (699 stars)

Publication unit: University of Illinois Champaign-Urbana Resource Library: https://github.com/hkchengrex/XMem Paper: https://arxiv.org/abs/2207.07115v2

8、 TinyViT: Fast Pretraining Distillation for Small Vision Transformers (656 stars)

Publisher: Microsoft Research Institute, Microsoft Cloud AI Resource Library: https://github.com/microsoft/cream Paper: https://arxiv.org/abs/2207.10666v1

9 , Towards Grand Unification of Object Tracking (644 stars)

Issuing unit: Dalian University of Technology, Byte Beat, University of Hong Kong, Pengcheng Laboratory

Resource library: ttps://github.com/masterbin-iiau/unicorn Paper: https://arxiv.org/abs/2207.07078 v3

10、Multiface: A Dataset for Neural Face Rendering（337 stars）

Отдел выпуска: Библиотека ресурсов Лаборатории метареальности: https://github.com/facebookresearch/multiface Документ: https://arxiv.org/abs/2207.11243v1

The above is the detailed content of Github's hot AI project list for July has a sequel! Nuwa Unlimited Edition can only be ranked third?. For more information, please follow other related articles on the PHP Chinese website!