機(jī)器之心報(bào)道
編輯:蛋醬
2021即將結(jié)束了,你今年讀了多少論文?
雖然世界仍在從新冠疫情的破壞中復(fù)蘇,人們無法向從前那樣時(shí)常線下相聚、共同探討交流關(guān)于學(xué)術(shù)領(lǐng)域的最新問題,但AI研究也沒有停下躍進(jìn)的步伐。
轉(zhuǎn)眼就是2021年底了,一年就這么就過去了,時(shí)光好像被偷走一樣。細(xì)細(xì)數(shù)來,你今年讀了多少論文?
一名加拿大博主Louis Bouchard以發(fā)布時(shí)間為順序,整理出了近40篇2021年不可錯(cuò)過的優(yōu)秀論文。整體來看,合集中的論文偏重計(jì)算機(jī)視覺方向。
在這個(gè)15分鐘左右的視頻中,你可以快速瀏覽這些論文:
以下是每篇論文的詳細(xì)信息:
1、DALLE: Zero-Shot Text-to-Image Generation from OpenAI
論文鏈接:https://arxiv.org/pdf/2102.12092.pdf
代碼地址:https://github.com/openai/DALL-E
視頻解讀:https://youtu.be/DJToDLBPovg
2、VOGUE: Try-On by StyleGAN Interpolation Optimization
論文鏈接:https://vogue-try-on.github.io/static_files/resources/VOGUE-virtual-try-on.pdf
視頻解讀:https://youtu.be/i4MnLJGZbaM
3、Taming Transformers for High-Resolution Image Synthesis
論文鏈接:https://compvis.github.io/taming-transformers/
代碼地址:https://github.com/CompVis/taming-transformers
視頻解讀:https://youtu.be/JfUTd8fjtX8
4、Thinking Fast And Slow in AI
論文鏈接:https://arxiv.org/abs/2010.06002
視頻解讀:https://youtu.be/3nvAaVSQxs4
5、Automatic detection and quantification of floating marine macro-litter inaerial images
論文鏈接:https://doi.org/10.1016/j.envpol.2021.116490
代碼地址:https://github.com/amonleong/MARLIT
視頻解讀:https://youtu.be/2dTSsdW0WYI
6、ShaRF: Shape-conditioned Radiance Fields from a Single View
論文鏈接:https://arxiv.org/abs/2102.08860
代碼地址:http://www.krematas.com/sharf/index.html
視頻解讀:https://youtu.be/gHkkrNMlGNg
7、Generative Adversarial Transformers
論文鏈接:https://arxiv.org/pdf/2103.01209.pdf
代碼地址:https://github.com/dorarad/gansformer
視頻解讀:https://youtu.be/HO-_t0UArd4
8、We Asked Artificial Intelligence to Create Dating Profiles. Would You Swipe Right?
論文鏈接:https://studyonline.unsw.edu.au/blog/ai-generated-dating-profile
代碼地址:https://colab.research.google.com/drive/1VLG8e7YSEwypxU-noRNhsv5dW4NfTGce#forceEdit=true&sandboxMode=true&scrollTo=aeXshJM-Cuaf
視頻解讀:https://youtu.be/IoRH5u13P-4
9、Swin Transformer: Hierarchical Vision Transformer using Shifted Windows
論文鏈接:https://arxiv.org/abs/2103.14030v2
代碼地址:https://github.com/microsoft/Swin-Transformer
視頻解讀:https://youtu.be/QcCJJOLCeJQ
10、IMAGE GANS MEET DIFFERENTIABLE RENDERING FOR INVERSE GRAPHICS AND INTERPRETABLE 3D NEURAL RENDERING
論文鏈接:https://arxiv.org/pdf/2010.09125.pdf
視頻解讀:https://youtu.be/dvjwRBZ3Hnw
11、Deep nets: What have they ever done for vision?
論文鏈接:https://arxiv.org/abs/1805.04025
視頻解讀:https://youtu.be/GhPDNzAVNDk
12、Infinite Nature: Perpetual View Generation of Natural Scenes from a Single Image
論文鏈接:https://arxiv.org/pdf/2012.09855.pdf
代碼地址:https://github.com/google-research/google-research/tree/master/infinite_nature
視頻解讀:https://youtu.be/NIOt1HLV_Mo
在線試用:https://colab.research.google.com/github/google-research/google-research/blob/master/infinite_nature/infinite_nature_demo.ipynb#scrollTo=sCuRX1liUEVM
13、Portable, Self-Contained Neuroprosthetic Hand with Deep Learning-Based Finger Control
論文鏈接:https://arxiv.org/abs/2103.13452
視頻解讀:https://youtu.be/wNBrCRzlbVw
14、Total Relighting: Learning to Relight Portraits for Background Replacement
論文鏈接:https://augmentedperception.github.io/total_relighting/total_relighting_paper.pdf
視頻解讀:https://youtu.be/rVP2tcF_yRI
15、LASR: Learning Articulated Shape Reconstruction from a Monocular Video
論文鏈接:https://openaccess.thecvf.com/content/CVPR2021/papers/Yang_LASR_Learning_Articulated_Shape_Reconstruction_From_a_Monocular_Video_CVPR_2021_paper.pdf
代碼地址:https://github.com/google/lasr
視頻解讀:https://youtu.be/lac7wqjS-8E
16、Enhancing Photorealism Enhancement
論文鏈接:http://vladlen.info/papers/EPE.pdf
代碼地址:https://github.com/isl-org/PhotorealismEnhancement
視頻解讀:https://youtu.be/3rYosbwXm1w
17、DefakeHop: A Light-Weight High-Performance Deepfake Detector
論文鏈接:https://arxiv.org/abs/2103.06929
視頻解讀:https://youtu.be/YMir8sRWRos
18、High-Resolution Photorealistic Image Translation in Real-Time: A Laplacian Pyramid Translation Network
論文鏈接:https://arxiv.org/pdf/2105.09188.pdf
代碼地址:https://github.com/csjliang/LPTN
視頻解讀:https://youtu.be/X7WzlAyUGPo
19、Barbershop: GAN-based Image Compositing using Segmentation Masks
論文鏈接:https://arxiv.org/pdf/2106.01505.pdf
代碼地址:https://github.com/ZPdesu/Barbershop
視頻解讀:https://youtu.be/HtqYMvBVJD8
20、TextStyleBrush: Transfer of text aesthetics from a single example
論文鏈接:https://arxiv.org/abs/2106.08385
代碼地址:https://github.com/facebookresearch/IMGUR5K-Handwriting-Dataset?fbclid=IwAR0pRAxhf8Vg-5H3fA0BEaRrMeD21HfoCJ-so8V0qmWK7Ub21dvy_jqgiVo
視頻解讀:https://youtu.be/hhAri5fl-XI
21、Animating Pictures with Eulerian Motion Fields
論文鏈接:https://arxiv.org/abs/2011.15128
代碼地址:https://eulerian.cs.washington.edu/
視頻解讀:https://youtu.be/KgTa2r7d0I0
22、CVPR 2021 Best Paper Award: GIRAFFE - Controllable Image Generation
論文鏈接:http://www.cvlibs.net/publications/Niemeyer2021CVPR.pdf
代碼地址:https://github.com/autonomousvision/giraffe
視頻解讀:https://youtu.be/JIJkURAkCxM
23、GitHub Copilot & Codex: Evaluating Large Language Models Trained on Code
論文鏈接:https://arxiv.org/pdf/2107.03374.pdf
代碼地址:https://copilot.github.com/
視頻解讀:https://youtu.be/az3oVVkTFB8
24、Apple: Recognizing People in Photos Through Private On-Device Machine Learning
論文鏈接:https://machinelearning.apple.com/research/recognizing-people-photos
視頻解讀:https://youtu.be/LIV-M-gFRFA
25、Image Synthesis and Editing with Stochastic Differential Equations
論文鏈接:https://arxiv.org/pdf/2108.01073.pdf
代碼地址:https://github.com/ermongroup/SDEdit
視頻解讀:https://youtu.be/xoEkSWJSm1k
https://colab.research.google.com/drive/1KkLS53PndXKQpPlS1iK-k1nRQYmlb4aO?usp=sharing
26、Sketch Your Own GAN
論文鏈接:https://arxiv.org/abs/2108.02774
代碼地址:https://github.com/PeterWang512/GANSketching
視頻解讀:https://youtu.be/vz_wEQkTLk0
27、Tesla's Autopilot Explained
在今年8月的特斯拉AI日上,特斯拉AI總監(jiān)Andrej Karpathy和其他人展示了特斯拉是如何通過八個(gè)攝像頭采集圖像,打造了基于視覺的自動(dòng)駕駛系統(tǒng)。
視頻解讀:https://youtu.be/DTHqgDqkIRw
28、Styleclip: Text-driven manipulation of StyleGAN imagery
論文鏈接:https://arxiv.org/abs/2103.17249
代碼地址:https://github.com/orpatashnik/StyleCLIP
視頻解讀:https://youtu.be/RAXrwPskNso
https://colab.research.google.com/github/orpatashnik/StyleCLIP/blob/main/notebooks/StyleCLIP_global.ipynb
29、TimeLens: Event-based Video Frame Interpolation
論文鏈接:http://rpg.ifi.uzh.ch/docs/CVPR21_Gehrig.pdf
代碼地址:https://github.com/uzh-rpg/rpg_timelens
視頻解讀:https://youtu.be/HWA0yVXYRlk
30、Diverse Generation from a Single Video Made Possible
論文鏈接:https://arxiv.org/abs/2109.08591
代碼地址:https://nivha.github.io/vgpnn/
視頻解讀:https://youtu.be/Uy8yKPEi1dg
31、Skillful Precipitation Nowcasting using Deep Generative Models of Radar
論文鏈接:https://arxiv.org/pdf/2110.09958.pdf
代碼地址:https://cocktail-fork.github.io/
視頻解讀:https://youtu.be/Rpxufqt5r6I
33、ADOP: Approximate Differentiable One-Pixel Point Rendering
論文鏈接:https://arxiv.org/pdf/2110.06635.pdf
代碼地址:https://github.com/darglein/ADOP
視頻解讀:https://youtu.be/Jfph7Vld_Nw
34、(Style)CLIPDraw: Coupling Content and Style in Text-to-Drawing Synthesis
CLIPDraw論文鏈接:https://arxiv.org/abs/2106.14843
在線試用:https://colab.research.google.com/github/kvfrans/clipdraw/blob/main/clipdraw.ipynb
StyleCLIPDraw論文鏈接:https://arxiv.org/abs/2111.03133
在線試用:https://colab.research.google.com/github/pschaldenpand/StyleCLIPDraw/blob/master/Style_ClipDraw.ipynb
視頻解讀:https://youtu.be/5xzcIzHm8Wo
35、SwinIR: Image restoration using swin transformer
論文鏈接:https://arxiv.org/abs/2108.10257
代碼地址:https://github.com/JingyunLiang/SwinIR
視頻解讀:https://youtu.be/GFm3RfrtDoU
https://replicate.ai/jingyunliang/swinir
36、EditGAN: High-Precision Semantic Image Editing
論文鏈接:https://arxiv.org/abs/2111.03186
代碼地址:https://nv-tlabs.github.io/editGAN/
視頻解讀:https://youtu.be/bus4OGyMQec
37、CityNeRF: Building NeRF at City Scale
論文鏈接:https://arxiv.org/pdf/2112.05504.pdf
代碼地址:https://city-super.github.io/citynerf/
視頻解讀:https://youtu.be/swfx0bJMIlY
38、ClipCap: CLIP Prefix for Image Captioning
論文鏈接:https://arxiv.org/abs/2111.09734
代碼地址:https://github.com/rmokady/CLIP_prefix_caption
視頻解讀:https://youtu.be/VQDrmuccWDo
在線試用:https://colab.research.google.com/drive/1tuoAC5F4sC7qid56Z0ap-stR3rwdk0ZV?usp=sharing
當(dāng)然,博主在整理的過程中也不能保證完美。經(jīng)網(wǎng)友提醒,這里可以手動(dòng)添加一項(xiàng)突破性研究:「AlphaFold」。
去年,谷歌旗下人工智能技術(shù)公司 DeepMind 宣布深度學(xué)習(xí)算法「Alphafold」破解了出現(xiàn)五十年之久的蛋白質(zhì)分子折疊問題。2021年7月,AlphaFold 的論文正式發(fā)表在《Nature》雜志上。