2020: A Year Full of Amazing AI Papers - A Review

A curated list of the latest breakthroughs in AI in 2020 by release date with a clear video explanation, link to a more in-depth article, and code

2020: A Year Full of Amazing AI Papers - A Review

Even with everything that happened in the world this year, we still had the chance to see a lot of amazing research come out. Especially in the field of artificial intelligence. More, many important aspects were highlighted this year, like the ethical aspects, important biases, and much more. Artificial intelligence and our understanding of the human brain and its link to AI is constantly evolving, showing promising applications in the soon future.

Here are the most interesting research papers of the year, in case you missed any of them. In short, it is basically a curated list of the latest breakthroughs in AI and Data Science by release date with a clear video explanation, link to a more in-depth article, and code (if applicable). Enjoy the read, and let me know if I missed any important papers in the comments, or by contacting me directly on LinkedIn!

The complete reference to each paper is listed at the end of this article.

Access the complete list in a GitHub repository

Tag me on Twitter (@Whats_AI) or LinkedIn (Louis (What’s AI) Bouchard) if you share the article!

Watch a complete 2020 rewind in 15 minutes

YOLOv4: Optimal Speed and Accuracy of Object Detection [1]

This 4th version has been recently introduced in April 2020 by Alexey Bochkovsky et al. in the paper “YOLOv4: Optimal Speed and Accuracy of Object Detection”. The main goal of this algorithm was to make a super-fast object detector with high quality in terms of accuracy.

Introduction to YOLOv4 | SOTA Real-Time Object Detection in 2020
The 4th version of this algorithm has just been released. And it has the best real-time object detection accuracy to date! It is essentially a real-time object detector, just like the initial YOLO…

Click here for the Yolo v4 code

DeepFaceDrawing: Deep Generation of Face Images from Sketches [2]

You can now generate high-quality face images from rough or even incomplete sketches with zero drawing skills using this new image-to-image translation technique! If your drawing skills as bad as mine you can even adjust how much the eyes, mouth, and nose will affect the final image! Let’s see if it really works and how they did it.

AI Generates Real Faces From Sketches!
You can now generate high-quality face images from rough or even incomplete sketches with zero drawing skills using this new image-to-image translation technique! If your drawing skills as bad as…

Click here for the DeepFaceDrawing code

Learning to Simulate Dynamic Environments with GameGAN [3]

PAC-MAN Recreated with AI by NVIDIA Researchers | NVIDIA Blog
On the 40th anniversary of PAC-MAN, NVIDIA GameGAN generated a fully functional version of the classic — without an underlying game engine.

Click here for GameGAN code

PULSE: Self-Supervised Photo Upsampling via Latent Space Exploration of Generative Models [4]

This new algorithm transforms a blurry image into a high-resolution image!
It can take a super low-resolution 16x16 image and turn it into a 1080p high definition human face! You don’t believe me? Then you can do just like me and try it on yourself in less than a minute! But first, let’s see how they did that.

This AI makes blurry faces look 60 times sharper
This new algorithm transforms a blurry image into a high-resolution image!It can take a super low-resolution 16x16 image and turn it into a 1080p high definition human face! You don’t believe me…

Click here for the PULSE code

Unsupervised Translation of Programming Languages [5]

This new model converts code from a programming language to another without any supervision! It can take a Python function and translate it into a C++ function, and vice-versa, without any prior examples! It understands the syntax of each language and can thus generalize to any programming language! Let’s see how they did that.

This AI translates code from a programming language to another | Facebook TransCoder Explained
This new model converts code from a programming language to another without any supervision! It can take a Python function and translate it into a C++ function, and vice-versa, without any prior…

Click here for the Transcoder code

PIFuHD: Multi-Level Pixel-Aligned Implicit Function for High-Resolution 3D Human Digitization [6]

This AI Generates 3D high-resolution reconstructions of people from 2D images! It only needs a single image of you to generate a 3D avatar that looks just like you, even from the back!

AI Generates 3D high-resolution reconstructions of people from 2D images | Introduction to PIFuHD
The coolest thing about this new paper is that they made a demo available on google colab where you can easily try it on yourself, as I will show in this article! But first, let’s see how they did…

Click here for the PiFuHD code

High-Resolution Neural Face Swapping for Visual Effects [7]

Researchers at Disney developed a new High-Resolution Face Swapping algorithm for Visual Effects in the paper of the same name. It is capable of rendering photo-realistic results at megapixel resolution. Working for Disney, they are most certainly the best team for this work. Their goal is to swap the face of a target actor from a source actor while maintaining the actor’s performance. This is incredibly challenging and is useful in many circumstances, such as changing the age of a character, when an actor is not available, or even when it involves a stunt scene that would be too dangerous for the main actor to perform. The current approaches require a lot of frame-by-frame animation and post-processing by professionals.

Disney’s New High Resolution Face Swapping Algorithm | New 2020 Face Swap Technology Explained
Researchers at Disney developed a new High-Resolution Face Swapping algorithm for Visual Effects in the paper of the same name. It is capable of rendering photo-realistic results at megapixel…

Swapping Autoencoder for Deep Image Manipulation [8]

This new technique can change the texture of any picture while staying realistic using complete unsupervised training! The results look even better than what GANs can achieve while being way faster! It could even be used to create deepfakes!

Texture-Swapping AI beats GANs for Image Manipulation!
This new technique can change the texture of any picture while staying realistic using complete unsupervised training! The results look even better than what GANs can achieve while being way faster…

Click here for the Swapping autoencoder code

GPT-3: Language Models are Few-Shot Learners [9]

The current state-of-the-art NLP systems struggle to generalize to work on different tasks. They need to be fine-tuned on datasets of thousands of examples while humans only need to see a few examples to perform a new language task. This was the goal behind GPT-3, to improve the task-agnostic characteristic of language models.

Can GPT-3 Really Help You and Your Company?
In this article, I will start by doing a short introduction to what is GPT-3. Then, we will see why and how GPT-3 can be useful to you and your company with many real-world application examples…

Click here for GPT-3’s GitHub page

Learning Joint Spatial-Temporal Transformations for Video Inpainting [10]

This AI can fill the missing pixels behind a removed moving object and reconstruct the whole video with way more accuracy and less blurriness than current state-of-the-art approaches!

This AI takes a video and fills the missing pixels behind an object!
This AI can fill the missing pixels behind a removed moving object and reconstruct the whole video with way more accuracy and less blurriness than current state-of-the-art approaches! Let’s see what…

Click here for this Video Inpainting code

Image GPT — Generative Pretraining from Pixels [11]

A good AI, like the one used in Gmail, can generate coherent text and finish your phrase. This one uses the same principles in order to complete an image! All done in an unsupervised training with no labels required at all!

This AI Can Generate the Other Half of a Picture Using a GPT Model
OpenAI recently shared a new paper called “Generative Pretraining from Pixels” which is used to predict pixels without incorporating knowledge of the 2-D image structure. They wanted to see if an…

Click here for the OpenAI’s Image GPT code

Learning to Cartoonize Using White-box Cartoon Representations [12]

This AI can cartoonize any picture or video you feed it in the cartoon style you want! Let’s see how it does that and some amazing examples. You can even try it yourself on the website they created as I did for myself!

This AI can cartoonize any picture or video you feed it! Paper Introduction & Results examples
This AI can cartoonize any picture or video you feed it in the cartoon style you want!Let’s see how it does that and some amazing examples. You can even try it yourself on the website they created…

Click here for the Cartoonize code

FreezeG: Freeze the Discriminator: a Simple Baseline for Fine-Tuning GANs [13]

This face generating model is able to transfer normal face photographs into distinctive styles such as Lee Mal-Nyeon’s cartoon style, the Simpsons, arts, and even dogs! The best thing about this new technique is that it’s super simple and significantly outperforms previous techniques used in GANs.

This Face Generating Model Transfers Real Face Photographs Into Distinctive Cartoon Styles |…
This face generating model is able to transfer normal face photographs into distinctive styles such as Lee Mal-Nyeon’s cartoon style, the Simpsons, arts, and even dogs! The best thing about this new…

Click here for the FreezeG code

Neural Re-Rendering of Humans from a Single Image [14]

The algorithm represents body pose and shape as a parametric mesh which can be reconstructed from a single image and easily reposed. Given an image of a person, they are able to create synthetic images of the person in different poses or with different clothing obtained from another input image.

Transfer clothes between photos using AI. From a single image!
The algorithm represents body pose and shape as a parametric mesh which can be reconstructed from a single image and easily reposed. Given an image of a person, they are able to create synthetic…

I2L-MeshNet: Image-to-Lixel Prediction Network for Accurate 3D Human Pose and Mesh Estimation from a Single RGB Image [15]

Their goal was to propose a new technique for 3D Human Pose and Mesh Estimation from a single RGB image. They called it I2L-MeshNet. Where I2L stands for Image-to-Lixel. Just like a voxel, volume + pixel, is a quantized cell in three-dimensional space, they defined lixel, a line, and pixel, as a quantized cell in one-dimensional space. Their method outperforms previous methods and the code is publicly available!

Accurate 3D Human Pose and Mesh Estimation from a Single RGB Image! With Code Publicly Avaibable!
With the ECCV conference coming soon, we have a ton of amazing papers coming out related to computer vision. Here’s an impressive one made by researchers at the Seoul National University about 3D…

Click here for the I2L-MeshNet code

Beyond the Nav-Graph: Vision-and-Language Navigation in Continuous Environments [16]

Language-guided navigation is a widely studied field and a very complex one. Indeed, it may seem simple for a human to just walk through a house to get to your coffee that you left on your nightstand to the left of your bed. But it is a whole other story for an agent, which is an autonomous AI-driven system using deep learning to perform tasks.

Language-Guided Navigation in a 3D Environment
Language-guided navigation is a widely studied field and a very complex one. Indeed, it may seem simple for a human to just walk through a house to get to your coffee that you left on your nightstand…

Click here for the VLN-CE code

RAFT: Recurrent All-Pairs Field Transforms for Optical Flow [17]

ECCV 2020 Best Paper Award Goes to Princeton Team. They developed a new end-to-end trainable model for optical flow. Their method beats state-of-the-art architectures’ accuracy across multiple datasets and is way more efficient. They even made the code available for everyone on their Github!

ECCV 2020 Best Paper Award | A New Architecture For Optical Flow
ECCV 2020 Best Paper Award Goes to Princeton Team.They developed a new end-to-end trainable model for optical flow.Their method beats state-of-the-art architectures’ accuracy across multiple…

Click here for the RAFT code

Crowdsampling the Plenoptic Function [18]

Using tourists’ public photos from the internet, they were able to reconstruct multiple viewpoints of a scene conserving the realistic shadows and lighting! This is a huge advancement of the state-of-the-art techniques for photorealistic scene rendering and their results are simply amazing.

Reconstruct Photorealistic Scenes from Tourists’ Public Photos on the Internet!
Using tourists’ public photos from the internet, they were able to reconstruct multiple viewpoints of a scene conserving the realistic shadows and lighting!This is a huge advancement of the…

Click here for the Crowdsampling code

Old Photo Restoration via Deep Latent Space Translation [19]

Imagine having the old, folded, and even torn pictures of your grandmother when she was 18 years old in high definition with zero artifacts. This is called old photo restoration and this paper just opened a whole new avenue to address this problem using a deep learning approach.

Old Photo Restoration using Deep Learning
Imagine having the old, folded, and even torn pictures of your grandmother when she was 18 years old in high definition with zero artifacts.This is called old photo restoration and this paper just…

Click here for the Old Photo Restoration code

Neural circuit policies enabling auditable autonomy [20]

Researchers from IST Austria and MIT have successfully trained a self-driving car using a new artificial intelligence system based on the brains of tiny animals, such as threadworms. They achieved that with only a few neurons able to control the self-driving car, compared to the millions of neurons needed by the popular deep neural networks such as Inceptions, Resnets, or VGG. Their network was able to completely control a car using only 75 000 parameters, composed of 19 control neurons, rather than millions!

A New Brain-inspired Intelligent System Drives a Car Using Only 19 Control Neurons!
Researchers are always looking for new ways to build intelligent models. We all know that really deep supervised models work great when we have sufficient data to train them, but one of the hardest…

Click here for the NCP code

Lifespan Age Transformation Synthesis [21]

A team of researchers from Adobe Research developed a new technique for age transformation synthesis based on only one picture from the person. It can generate the lifespan pictures from any picture you sent it.

Generate Younger & Older Versions of Yourself!
A team of researchers from Adobe Research developed a new technique for age transformation synthesis based on only one picture from the person.It can generate the lifespan pictures from any picture…

Click here for the Lifespan age transformation synthesis code

DeOldify [22]

DeOldify is a technique to colorize and restore old black and white images or even film footage. It was developed and is still getting updated by only one person Jason Antic. It is now the state of the art way to colorize black and white images, and everything is open-sourced, but we will get back to this in a bit.

This AI can Colorize your Black & White Photos with Full Photorealistic Renders! (DeOldify)
DeOldify is a technique to colorize and restore old black and white images or even film footage. It was developed and is still getting updated by only one person Jason Antic. It is now the state of…

Click here for the DeOldify code

COOT: Cooperative Hierarchical Transformer for Video-Text Representation Learning [23]

As the name states, it uses transformers to generate accurate text descriptions for each sequence of a video, using both the video and a general description of it as inputs.

Video to Text Description Using Deep Learning and Transformers | COOT
As many of you guys may already know, we are approaching the date of the Neural Information Processing Systems conference, also referred to as the NeurIPS conference. Where many awesome papers will…

Click here for the COOT code

Stylized Neural Painting [24]

This Image-to-Painting Translation method simulates a real painter on multiple styles using a novel approach that does not involve any GAN architecture, unlike all the current state-of-the-art approaches!

Image-to-Painting Translation With Style Transfer
One of the best things about this paper is that their code is completely available on GitHub and they even created two google colab notebooks for you to try on your own pictures! But first, let’s…

Click here for the Stylized Neural Painting code

Is a Green Screen Really Necessary for Real-Time Portrait Matting? [25]

Human matting is an extremely interesting task where the goal is to find any human in a picture and remove the background from it. It is really hard to achieve due to the complexity of the task, having to find the person or people with the perfect contour. In this post, I review the best techniques used over the years and a novel approach published on November 29th, 2020. Many techniques are using basic computer vision algorithms to achieve this task, such as the GrabCut algorithm, which is extremely fast, but not very precise.

High-Quality Background Removal Without Green Screens
Human matting is an extremely interesting task where the goal is to find any human in a picture and remove the background from it. It is really hard to achieve due to the complexity of the task…

Click here for the MODNet code

ADA: Training Generative Adversarial Networks with Limited Data [26]

With this new training method developed by NVIDIA, you can train a powerful generative model with one-tenth of the images! Making possible many applications that do not have access to so many images!

GAN Training Breakthrough for Limited Data Applications & New NVIDIA Program! NVIDIA Research
This new paper covers a technique for training a GAN architecture. They are used in many applications related to computer vision, where we want to generate a realistic transformation of an image…

Click here for the ADA code

Improving Data‐Driven Global Weather Prediction Using Deep Convolutional Neural Networks on a Cubed Sphere [27]

The current traditional approach for weather forecasting uses what we call “Numerical weather prediction” models. It uses mathematical models of the atmosphere and oceans to predict the weather based on the current conditions. It was first introduced in the 1920s and produced realistic results in the 1950s using computer simulations. These mathematical models work for predicting both short and long-term forecasts. But it’s heavy in computation and cannot base its predictions on as much data as a deep neural network. This is partly why it is so promising. These current numerical weather prediction models already use machine learning to improve the forecasts as a post-processing tool. Weather forecasting is receiving more and more attention from machine learning researchers, already yielding promising results.

AI is Predicting Faster and More Accurate Weather Forecasts
The current traditional approach for weather forecasting uses what we call “Numerical weather prediction” models. It uses mathematical models of the atmosphere and oceans to predict the weather based…

Click here for the weather forecasting code

NeRV: Neural Reflectance and Visibility Fields
for Relighting and View Synthesis [28]

This new method is able to generate a complete 3-dimensional scene and has the ability to decide the lighting of the scene. All this with very limited computation costs and amazing results compared to previous approaches.

Generate a Complete 3D Scene Under Arbitrary Lighting Conditions from a Set of Input Images
NeRV, or Neural Reflectance and Visibility Fields for Relighting and View Synthesis, is a method that produces a 3D representation of a scene and can generate arbitrary lighting conditions. It only…

Click here for the NeRV code (available soon)

Conclusion

As you can see, this was an extremely insightful year for artificial intelligence, and I am super excited to see what’s going to happen in 2021! I will be sure to cover the most exciting and interesting papers, and I would love it if you could take part in this adventure! If you like my work and want to stay up-to-date with AI, you should definitely follow me on my other social media accounts (LinkedIn, Twitter) and subscribe to my weekly AI newsletter!

To support me:

  • The best way to support me is by following me here on Medium or subscribe to my channel on YouTube if you like the video format.
  • Support my work on Patreon
  • Join our Discord community: Learn AI Together and share your projects, papers, best courses, find Kaggle teammates, and much more!

Access the complete list in a GitHub repository

Tag me on Twitter (@Whats_AI) or LinkedIn (Louis (What’s AI) Bouchard) if you share the article!

If you are interested in AI research, here is another great article for you:

Top 10 Computer Vision Papers 2020
Even with everything that happened in the world this year, we still had the chance to see a lot of amazing research come out. Especially in the field of artificial intelligence and more precisely…

Paper references

[1] A. Bochkovskiy, C.-Y. Wang, and H.-Y. M. Liao, Yolov4: Optimal speed and accuracy of object detection, 2020. arXiv:2004.10934 [cs.CV].

[2] S.-Y. Chen, W. Su, L. Gao, S. Xia, and H. Fu, “DeepFaceDrawing: Deep generation of face images from sketches,” ACM Transactions on Graphics (Proceedings of ACM SIGGRAPH2020), vol. 39, no. 4, 72:1–72:16, 2020.

[3] S. W. Kim, Y. Zhou, J. Philion, A. Torralba, and S. Fidler, “Learning to Simulate DynamicEnvironments with GameGAN,” in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Jun. 2020.

[4] S. Menon, A. Damian, S. Hu, N. Ravi, and C. Rudin, Pulse: Self-supervised photo upsampling via latent space exploration of generative models, 2020. arXiv:2003.03808 [cs.CV].

[5] M.-A. Lachaux, B. Roziere, L. Chanussot, and G. Lample, Unsupervised translation of programming languages, 2020. arXiv:2006.03511 [cs.CL].

[6] S. Saito, T. Simon, J. Saragih, and H. Joo, Pifuhd: Multi-level pixel-aligned implicit function for high-resolution 3d human digitization, 2020. arXiv:2004.00452 [cs.CV].

[7] J. Naruniec, L. Helminger, C. Schroers, and R. Weber, “High-resolution neural face-swapping for visual effects,” Computer Graphics Forum, vol. 39, pp. 173–184, Jul. 2020.doi:10.1111/cgf.14062.

[8] T. Park, J.-Y. Zhu, O. Wang, J. Lu, E. Shechtman, A. A. Efros, and R. Zhang,Swappingautoencoder for deep image manipulation, 2020. arXiv:2007.00653 [cs.CV].

[9] T. B. Brown, B. Mann, N. Ryder, M. Subbiah, J. Kaplan, P. Dhariwal, A. Neelakantan, P.Shyam, G. Sastry, A. Askell, S. Agarwal, A. Herbert-Voss, G. Krueger, T. Henighan, R. Child, A. Ramesh, D. M. Ziegler, J. Wu, C. Winter, C. Hesse, M. Chen, E. Sigler, M. Litwin, S.Gray, B. Chess, J. Clark, C. Berner, S. McCandlish, A. Radford, I. Sutskever, and D. Amodei,“Language models are few-shot learners,” 2020. arXiv:2005.14165 [cs.CL].

[10] Y. Zeng, J. Fu, and H. Chao, Learning joint spatial-temporal transformations for video in-painting, 2020. arXiv:2007.10247 [cs.CV].

[11] M. Chen, A. Radford, R. Child, J. Wu, H. Jun, D. Luan, and I. Sutskever, “Generative pretraining from pixels,” in Proceedings of the 37th International Conference on Machine Learning, H. D. III and A. Singh, Eds., ser. Proceedings of Machine Learning Research, vol. 119, Virtual: PMLR, 13–18 Jul 2020, pp. 1691–1703. [Online]. Available:http://proceedings.mlr.press/v119/chen20s.html.

[12] Xinrui Wang and Jinze Yu, “Learning to Cartoonize Using White-box Cartoon Representations.”, IEEE Conference on Computer Vision and Pattern Recognition, June 2020.

[13] S. Mo, M. Cho, and J. Shin, Freeze the discriminator: A simple baseline for fine-tuning gans,2020. arXiv:2002.10964 [cs.CV].

[14] K. Sarkar, D. Mehta, W. Xu, V. Golyanik, and C. Theobalt, “Neural re-rendering of humans from a single image,” in European Conference on Computer Vision (ECCV), 2020.

[15] G. Moon and K. M. Lee, “I2l-meshnet: Image-to-lixel prediction network for accurate 3d human pose and mesh estimation from a single rgb image,” in European Conference on ComputerVision (ECCV), 2020

[16] J. Krantz, E. Wijmans, A. Majumdar, D. Batra, and S. Lee, “Beyond the nav-graph: Vision-and-language navigation in continuous environments,” 2020. arXiv:2004.02857 [cs.CV].

[17] Z. Teed and J. Deng, Raft: Recurrent all-pairs field transforms for optical flow, 2020. arXiv:2003.12039 [cs.CV].

[18] Z. Li, W. Xian, A. Davis, and N. Snavely, “Crowdsampling the plenoptic function,” inProc.European Conference on Computer Vision (ECCV), 2020.

[19] Z. Wan, B. Zhang, D. Chen, P. Zhang, D. Chen, J. Liao, and F. Wen, Old photo restoration via deep latent space translation, 2020. arXiv:2009.07047 [cs.CV].

[20] Lechner, M., Hasani, R., Amini, A. et al. Neural circuit policies enabling auditable autonomy. Nat Mach Intell 2, 642–652 (2020). https://doi.org/10.1038/s42256-020-00237-3

[21] R. Or-El, S. Sengupta, O. Fried, E. Shechtman, and I. Kemelmacher-Shlizerman, “Lifespanage transformation synthesis,” in Proceedings of the European Conference on Computer Vision(ECCV), 2020.

[22] Jason Antic, Creator of DeOldify, https://github.com/jantic/DeOldify

[23] S. Ging, M. Zolfaghari, H. Pirsiavash, and T. Brox, “Coot: Cooperative hierarchical trans-former for video-text representation learning,” in Conference on Neural Information ProcessingSystems, 2020.

[24] Z. Zou, T. Shi, S. Qiu, Y. Yuan, and Z. Shi, Stylized neural painting, 2020. arXiv:2011.08114[cs.CV].

[25] Z. Ke, K. Li, Y. Zhou, Q. Wu, X. Mao, Q. Yan, and R. W. Lau, “Is a green screen really necessary for real-time portrait matting?” ArXiv, vol. abs/2011.11961, 2020.

[26] T. Karras, M. Aittala, J. Hellsten, S. Laine, J. Lehtinen, and T. Aila, Training generative adversarial networks with limited data, 2020. arXiv:2006.06676 [cs.CV].

[27] J. A. Weyn, D. R. Durran, and R. Caruana, “Improving data-driven global weather prediction using deep convolutional neural networks on a cubed sphere”, Journal of Advances in Modeling Earth Systems, vol. 12, no. 9, Sep. 2020, issn: 1942–2466.doi:10.1029/2020ms002109

[28] P. P. Srinivasan, B. Deng, X. Zhang, M. Tancik, B. Mildenhall, and J. T. Barron, “Nerv: Neural reflectance and visibility fields for relighting and view synthesis,” in arXiv, 2020.