Gather Town: Virtual space meets video meeting December 15, 2020 - Wav2Vec 2.0 Revealed-Create ASR with 10 Minute Voice August 5, 2020
- Pororo - KakaoBrain's Integrated Natural Language Framework February 10, 2021
- Generation code of Lee Malnyun webtoon style faces September 21, 2020
- Open source chatbot frameworks: Kochat, Rasa, Rocket Chat December 30, 2020
News
Smilegate Hope Studio delivers donations raised with SeAH to Haemil School
A message of hope from SeAH Story... Delivered a total of 30 million won including 24-hour donation concerts Hope Studio, game users-'Pleasant donation' with viewers of internet broadcasting
Smilegate'SeAH' challenged 24-hour relay donation concert to'make the world beautiful'
Schedule full of 24 hours from noon on the 29th… A surprise guest appearance is also announced! Full donation for broadcasting, delivered to Hope Studio... Multicultural alternative educational institution'Haemil School' IT infrastructure support...
Smilegate's mentoring to return as much as you learn
In the popular drama <Startup>, young people dreaming of starting a business appear. Behind those who are full of passion, the first challenges such as business, organization, etc.
Smilegate.AI facebook page opened
Smilegate.AI's facebook page has been opened. On this homepage, mainly technical and professional contents are shared, while on the facebook page, news about the center, research process, and life…
Smilegate Future Lab supports AI text interpretation service for children and adolescents with hearing impairments
AI-based real-time text interpretation software service support MAP tool pack for children and adolescents with hearing impairments in the blind spot in the online education environment caused by Corona 19…
Rainbow project, a creative community that grows together
'Creativity' is a staple item that has not been left out of the talent perspective of recent companies. To become a leader rather than a follower in a fast-changing era...
Weekly Popular
Recruit
Hiring AI Visual Experts
Among AI fields, GAN, Style Transfer, Retargeting, etc. Visual-related technologies are widely understood, and from this, we will propose a project and lead the direction…
Hiring Android App Developers
We are looking for proactive talents with mobile app development capabilities and the ability to Fast Prototyping new ideas. Various AI technologies...
Hiring AI experts
We look for active talents who have a broad understanding of the field of AI and the latest technologies, and can propose projects and lead the direction from them. Innovative…
Service Developer Hire
We are looking for active talents who have the ability to develop new ideas in various fields and fast prototyping new ideas. Various AI technologies...
Recent Posts
SPRi Metaverse Report
Metaverse, which means all forms of interaction between reality and connected avatars within a virtual environment, has recently attracted attention from various industries. Especially…
GPT-Neo: Open Source GPT-3 Project
OpenAI's GPT-3 is a large language model with a parameter count of up to 175B. Despite the amazing results that GPT-3 shows, it's not open sourced...
Applications of Graph Neural Network
Graph Neural Network (GNN) is emerging as a new field of AI. GNN is when data is expressed in a graph, which is a structure connected by a number of nodes and edges,...
ImageNet and privacy
ImageNet is a dataset that has greatly influenced the advancement of AI technology so that no one knows about AI researchers. A large number of images and about...
Analyzable control method of GAN model (GANSpace)
As a deep learning-based image generation method, GAN produces a lot of amazing results. In particular, by changing the latent vector after learning, a number of meaningful meanings...
$40M Series B of Hugging Face
It is news that Hugging Face, famous for its deep learning-based natural language processing open source, has attracted a $40M Series B investment. Hugging Face Transformer is probably a natural language...
A collection of AI projects for mobile devices (Awesome Tensorflow Lite)
Tensorflow Lite is a software package that contains tools that allow AI models trained with Tensorflow to run on mobile devices. Currently 4 billion...
Jina - open source multi-modal search engine based on deep learning
Jina, which is open sourced by Jina.AI, is a multimodal data search engine using deep learning technology. Not only some functions for search are implemented, but in the service...
Artificial Intelligence and Games (Book)
The latest AI technologies based on deep learning are being actively introduced in various fields such as speech recognition, speech synthesis, translation, chatbots, and smart factories, but games...
Google Lyra - speech compression based on deep generative models
Google Lyra is a new speech compression method based on the generative model. Existing speech compression methods achieve the original level of sound quality, that is, transparent quality...
AI market size in 2021 is predicted to be about $328B (IDC)
IDC, renowned for its market analysis in various sectors, has published predictions about the size of the AI market. Below is the relevant link: According to the announcement, in 2021…
Avatarify - Add real-time avatar to video communication
Avatarify is a program that adds real-time avatar animation functions to various video communication programs such as Zoom, Teams, Hangout, and Skype, and is open source. Camera of video communication program...
DensePhrases - Near-Real-Time Wikipedia Open Domain Q&A
DensePhrases is an open domain Q&A technology created by Jinhyuk Lee at Korea University, and was published as a paper titled “Learning Dense Representations of Phrases at Scale”. Here is a link to the paper:…
Framework merging of natural language processing and speech recognition
HuggingFace, famous for its integrated natural language processing package, adds speech recognition. The following is the related link: Specifically, Wav2Vec 2.0 developed by Facebook was added, Wav2Vec…
MeInGame-Create 3D models from photos
The technology to create 3D models from a single photo has been unveiled under the name MeInGame. Looking at the results posted on the public repository, the designer's work...
PapersWithCode's Korean dataset
Paperswithcode provides information on various papers in the field of AI, linked open sources, and SOTA, with links to over 3,000 useful datasets...
Pororo - KakaoBrain's Integrated Natural Language Framework
In Kakao Brain, Pororo, an integrated natural language framework capable of responding to various natural language tasks, has been released as open source. Pororo is the Platform Of neuRal mOdels for natuRal…
Ubuntu Dialog Corpus
Building a dialogue system that allows humans to have natural-feeling conversations with virtual agents is a difficult task in natural language processing, and a lot of ongoing research...
Digital Twin + AI
Digital Twin means moving an object or environment that has a substance in the real world into a virtual space and linking the two. Simply put…
RepVGG: a new interpretation of the Residual Network
Since the advent of AlexNet consisting of multiple convolution layers, there have been many studies on the structure of deep learning models. For example, Google Inception is 3×3,…
Memory optimization technology for very large models
As the number of parameters of a deep learning model increases significantly, the memory required for training is also increasing. OpenAI's GPT-2 consists of 1.5B parameters, and Google's mT5 is also in 13B...
Distributed training framework: Horovod and RaySGD
As deep learning models grow exponentially in size, it is no longer difficult to achieve usable learning times with a single machine. Well-known conversation...
Five brain functions to consider in AI research
It is true that the field of AI has made a lot of progress, but there are still many shortcomings compared to humans. What if the ultimate goal of AI is the human brain...
OpenAI DALL-E: Creating images from text
DALL-E, released by OpenAI, is a technology that generates images from natural language text. Previously, there were technologies for the same purpose such as StackGAN and OP-GAN, but DALL-E is a very large language model...
DeBERTa: surpass human performance in SuperGLUE
SuperGLUE is a challenge that evaluates the performance of AI technologies for a variety of natural language understanding tasks. It is composed of tasks with relatively high difficulty compared to existing GLUE...
What is the next step of AI in 2021?
Since deep learning began in earnest in 2012, AI technology has surpassed the performance of existing technologies in many fields. Although it is a limited environment...
Open-domain chatbot 'Luda' parenting diary: record from birth to close-beta
“Luda” of Scatterlab (https://scatterlab.co.kr/), whose official version was recently released, is an open domain chatbot learned based on billions of KakaoTalk conversation data. Anyone can chat through Facebook messenger. …
Technology to diagnose Alzheimer's with verbal information
Dementia is a phenomenon in which the brain function is greatly degraded to the point that it interferes with daily life. Alzheimer's is the most responsible for 60%-80% among dementia...
Open source chatbot frameworks: Kochat, Rasa, Rocket Chat
KoChat is a Korean open source chatbot framework released by Hyunwoong Ko. Here is the KoChat github repository: When we talk about chatbots, we often only think of the conversation model, but in fact…
FrankMoCap: 3D body and hand pose estimation technology as an alternative to motion capture
FrankMocap, a technology released by Facebook AI Research (FAIR), is responsible for extracting a pose for a 3D model from a single image or video. In particular, the body…
Digital Human Platform Companies
The Digital Human Platform is a form that combines various AI technologies with an avatar with a humanoid appearance. Thanks to advances in AI dialogue technology and visualization technology...
AI trends and game application examples
This is a report that summarizes AI trends and cases of AI technology application by game companies. The approximate table of contents is as follows: AI is… AI market and major…
Korean profanity text dataset
We share a set of Korean profanity data collected and labeled by Joonhee Jo. It is gathered from multiple communities, and seems to be suitable for evaluation of real-world data. Below is...
Can BERTology understand language?
The large-scale language model based on deep learning represented by BERT excels in various tasks related to natural language such as Q&A, document summary, document generation, and conversation...
Gather Town: Virtual space meets video meeting
Gather Town is a kind of video meeting solution such as Zoom or Teams, but it is characterized by actively introducing virtual space and avatars. For example, a virtual space called “Office”…
UnifiedQA: A single model responds to multiple Q&A tasks
QA Tasks that generate appropriate answers to a given question have seen a lot of performance gains due to recent deep learning technologies. The well-known SQuAD is also...
Techniques for generating questions from paragraphs
A problem commonly referred to as a Q&A task is to learn from a data set recorded in pairs of questions and answers so that when a question is asked, an appropriate answer comes out...
POSTECH STUDIOGAN: GAN algorithm library
StudioGAN is a pytorch-based open source library released by Pohang University CVLab Kang Min-guk, and various GAN algorithms are implemented. Included GAN algorithms include DCGAN, LSGAN, WGAN…
FACEBOOK REBEL showing poker skills beyond humans
It is not an exaggeration to say that poker is half a psychological game, so it is a different game from Go or chess. ReBeL released by Facebook this time is remarkable in this respect...
MELD: Multimodal EmotionLines Dataset
Multimodal EmotionLines Dataset (MELD) is a multimodal extension of EmotionLines, an emotionally labeled dialogue data set. MELD is what EmotionLines can use...
JALI's face animation technology used in CYBERPUNK 2077
Cyberpunk 2077, set to launch in late 2020 by CD PROJEKT RED studio, famous for the Witcher series, uses JaliResearch's facial animation technology. The main purpose is 3D…
MindMeld Conversational AI Platform
MindMeld is an open source interactive AI platform designed to ensure serviceable quality. Written in Python, the latest NLP skills and knowledge…
STATE OF AI REPORT 2020
This is the State of AI Report 2020, a report that analyzes various changes in the AI field. This report is with AI investor Nathan Benaich...
Improved NPC AI of The Division 2
Division 2 is an online action RPG developed by Massive Entertainment and published by Ubisoft, set in Washington, DC, where smallpox is popular. Gamers with government agents...
Avatar technology entering the K-POP market
Avatar has been used in various forms such as SNS, customer response, and character expression in games long before the advent of AI technology.
Unity ArtEngine
Unity's ArtEngine is a tool that makes it easy to create high-quality graphic resources using AI-based technology. In Unity, these technologies are called AI-assisted artistry...
Facebook Denoiser: real-time speech enhancement
We share a link to denoiser's github, Facebook's real-time noise reduction technology that was announced at Interspeech 2020. It is implemented in Pytorch and the title of the original paper is “Real Time…
AI market size compared to smartphones
According to the IDC forecast report, the AI market size in 2020 is predicted to be about 157B$. Of course, this number is in various industries related to AI, namely...
Video QA – 3D Attention is All You Need
Typically, Q&A systems use text to answer questions. With this type of task, you give a paragraph explaining a fact, ask a question, and give an appropriate answer...
UneeQ launches digital human platform
UneeQ has launched a digital human platform called Digital Human Creator. Although the service price is a bit burdensome, we offer a free trial, so we do a simple test…
VIBE: predicting 3D human model parameters from images
In order to implement the visualization part of Human-Like AI, it is necessary to think about how to create and move 3D human models, but various existing approaches are…
Multilingual translation model and language model trained with data from over 100 countries
Many attempts are being made to expand the language model and translation model, which were previously studied mainly in English, into multiple languages. Google's mT5 is the original T5 (text-to-text…
AI technology that predicts Covid-19 infections through cough sounds
Corona 19 has yet to show signs of calming down worldwide. MIT has created an AI model that can check whether COVID-19 is infected from the cough sound recorded with a mobile phone...
Adobe Neural Filter: Changing the image editing paradigm
Adobe announced an AI-based editing tool called neural filter. Some say it's already included in the latest version of Photoshop. In the example function, the picture...
AI trends in media compression by four events in 2020
2020 is likely to be the first year for the application of AI technology in the field of media compression to be considered in earnest. Here's a brief look at the four events that took place this year…
Conversation design for open domain chatbots
On the Ping-Pong blog, there was an article titled "Conversation composition of Luda dreaming of a superhuman AI", but there are a number of parts to be considered when designing an open domain chatbot...
Bluetooth-based COVID-19 risk group identification technology
There are a number of studies related to Corona 19 using AI technology. The paper shared below is a study by Fraunhofer HHI published in Nature, from Bluetooth Low Energy (BLE)…
NVidia Maxine: AI-based video communication platform
NVidia unveiled a cloud-based video communication platform called Maxine. Maxine's feature is the full introduction of AI technology, specifically facial images such as H.264...
Super resolution and facial expression of Gwan-soon Yu's old photo
This is a picture that made me feel very salty. The picture of Yu Gwan-soon, which remains only in low quality, is restored in high quality and added with a smile. Deep learning-based face editing technology...
Vid2Player: video analysis-based tennis player motion generation
There seem to be a lot of technologies recently to create new motions by extracting motion from human motion. (vid2vid, vid2game, pose2pose) Vid2Player was researched at Stanford University,...
MEAL v2: Achieving ImageNet Top-1 80% with ResNet-50
When multiple network models are combined into an ensemble, performance increases. Since the total network size and inference time also increase, it is difficult to apply it in practice.
HuggingFace Datasets 1.0
The first stable version 1.0 of the Huggingface Datasets library has been released, making it easy to use NLP datasets and evaluation metrics. Now…
Scalable character animation based on reinforcement learning
It is natural to see virtual characters and moves reasonably in terms of the laws of physics, that is, human-like.
Wav2Lip: generate lip motion from voice
LipGAN is a technology that generates the shape of the lips of a face image using a voice signal, and when it is actually applied to a video, it is somewhat regrettable in terms of visual artifacts and naturalness of movement...
NLP acceleration with HuggingFace and ONNX Runtime
The performance improvement shown by Transformer-based language models is surprising, but as the model size increases exponentially, concerns about service costs are also becoming important. Bert-base or GPT-2…
Korean language corpus of National Institute of Korean Language
The National Institute of the Korean Language has released Korean language materials for artificial intelligence learning on a large scale (13 kinds of 1.8 billion words). It was built by solving the copyright problem, and created an online agreement on the'Everyone's Corpus' site,…
Super-human AI for Gran Turismo
The link is a review of a paper published by Sony and ETH Zurich, which exceeds human records by applying reinforcement learning to the famous car game Gran Turismo...
GPT-3 examples and minGPT project
Scatterlab (https://scatterlab.co.kr/), which is prominent in everyday conversational research, is an article on the Ping-Pong team blog.
Generation code of Lee Malnyun webtoon style faces
bryandlee's github has the results of image translation application using deep generative model and related research made into a webcomic in the late years of calm man. Research title...
Facebook TransCoder: Unsupervised Programming Language Translator
There have been many attempts to convert code written in one programming language into another, and there are many types of commercial tools.
Performance analysis of human and AI for image classification
Imagenet-1K (1000 class image classification problem) is a task that has been optimized with the development of CNN. AlexNet's TOP-5 error that announced the beginning of the deep learning era is about…
Necessity of interaction with AR Glass concept video
This is an AR Glass concept video created by a designer named Iskander Utebayev. Considering the concept video, it is quite fancy and once implemented, the Human-Machine Interface using smart devices...
Lip2Wav: Generates a voice signal from silent lip movement
I've heard stories that you can know what you're talking about with just the movement of your lips if you get special training.
KcBERT: Korean language model reflecting comments and new words
In the case of a large-scale language model, there was always a difficulty because there was no Korean model. Following SKT's KoBERT, Naver comment data, new words, etc...
Pixar's Super Resolution Technology and Its Applications
Deep learning-based super resolution technology was adopted under the name DLSS (deep learning super sampling) in NVidia's latest GPU, and it has become a technology that is actually serviced to consumers.
Implementation feasibility with Google MixNet
The convolution commonly used in images is a 3D operation. (KxKxC; K=kernel size, C=number of channels) After applying this by dividing it into multiple 2D operations of KxKx1, 1x1xC in the channel direction...
Creating body movements by voice
LipGan is the study of creating mouth shapes from speech signals. It is a technique that can be usefully used to create a virtual character's mouth animation, but when applied in practice...
Microsoft Teams Together mode
As non-face-to-face video meetings have become commonplace in recent years, more and more people use Zoom to conduct multi-person video conferences. A phenomenon called zoom fatigue is also attracting attention...
H.266/VVC standard and deep learning technology
An international standard for a new video codec named (ISO MPEG) VVC or (ITU-T) H.266 has been released. Share related articles. Deep learning technology...
AI Fall or Renaissance
According to various statistics, the number of AI-related major conference participants increased 6 times over 5 years, and the number of AI startups was 28% for 4 years…
Multimodal Q&A – Visual Dialog Task
The Visual Dialog task is a multimodal task that adds an image to a Q&A task that consists of a question and answer. For example, a white cat and a black dog together...
AI and human collaboration: a new collective intelligence
Share what you've recently enjoyed reading. In this article, humans and AI do well in different fields, and rather than one side replacing the other...
Motion Retargeting from Motion, Skeleton and Angle
We share the project page of “Learning Character-Agnostic Motion for Motion Retargeting in 2D”, the paper published at SIGGRAPH 2019. This paper (which can be different)…
Adobe Mixamo: 3D character model open data
In the game production side, we share a link to the Adobe Mixamo site that is already used a lot. When you enter, 121 3D characters and 2484 character motions come up...
FastSpeech2 Open Source
TensorflowTTS, an open source based on Tensorflow 2 that supports several latest TTS models such as Tacotron2, MelGan, FastSpeech, etc., has finally begun supporting Microsoft FastSpeech2. FastSpeech2 is a Transformer…
AI: Intelligence vs. Automation
Links are articles that cause a lot of concerns about the difference between Intelligence and Automation. Artificial Intelligence is known as a term that came into use when neural networks appeared in the 1960s...
Emotion recognition reflecting facial expressions and body movements
There have been various attempts to recognize emotions from images or images. It is provided in the cloud API and is known so much that it becomes a topic on SNS (joy 95%, etc.).
GAN-based Image Compression
In the field of video compression, there are also things like Moore's Law (the number of transistors doubles every two years), MPEG-1 in 1993, MPEG-4/AVC (H.264) in 2003, MPEG-H/HEVC in 2013 ( H.265)...
Text-to-SQL: Convert natural language to SQL
Text-to-SQL is a task that automatically converts natural language into SQL. The post I shared at the bottom was written by Aerin Kim of Microsoft, and it is well organized about Text-to-SQL.
Introducing NVidia Ampere Architecture
When learning deep learning algorithms, GPU is considered essential, but when serving after model training is complete, CPU is used instead of GPU...
Speech2Face-face prediction from speech signals
MIT's Speech2Face is a study that generates a speaker's face from a speech signal. However, it is not that speech to face transform is performed with one model, and it is an existing model for other purposes...
Google MixIT AI-Separation of unsupervised learning sound sources
MixIT AI, released by Google, is a technology that obtains a separate sound source from single-channel audio in which multiple sound sources are mixed. It can be viewed as a blind source separation task...
Algorithm aversion and explainable AI
In the field of prediction, Algorithm Aversion means that when you realize that an algorithm can make mistakes, you tend not to use it even if it is superior to human predictions...
Wav2Vec 2.0 Revealed-Create ASR with 10 Minute Voice
Facebook's wav2vec, which became a hot topic because it made a speech recognizer with only 10 minutes of labeled data after representation training with 53,000 hours of label-free data.
MIT DriveSeg-data for road situation awareness research
It is a dataset DriveSeg created for research on road situation awareness (used for self-driving cars, etc.). For each frame of the video, the entire image is pixel-by-pixel semantic labeling…
Introduction of autonomous vehicle technology and social consensus
Although it is a little leap forward, if you consider the addition of physical devices to AI algorithms as intelligent robots, the intelligent robot that will be most popular in the future is...
Human brain vs. AI-Hardware Comparison
One of the recent trends is the super-giant model, i.e. the enormous increase in the number of parameters and the application of traditional learning methods. The “software capability” of the human brain…
Machine reading comprehension (MRC) task and data set arrangement
Many MRC models proposed so far show evaluation values beyond human capabilities in various tasks and datasets, but better than humans for a given context...
IBM's emotional robot Nao-mi
This is a video of IBM's emotional robot Nao-mi. [Summary of Contents] A robot that says he doesn't want to do with a person who asks to destroy a tower that has been difficult to build. To the continual demand...
Transfer Learning becomes a necessity, not an option
The learning cost of GPT-3, a pronoun of the super-scale language model, which showed the possibility that it can be applied to all natural language tasks only with Few shot learning, is about Hanwha...
First Order Motion Model for Image Animation
Los Angeles Noir, a 2011 film made by Rockstar, surprised many with facial animations that were far superior to other games. The technology used at this time...
The phenomenon of knowledge unindex by YouTube's advancement
With the advent of digitalization and the advent of the Internet and the web, knowledge is distributed and stored on servers around the world, connected to each other, and made searchable, so that accessibility and usability are dramatically improved. Books…
Codec Avatar on Facebook
A demo video of Facebook's digital human project under the name “Codec Avatar” has been released. This is an added part compared to the 2019 video, and the avatar looks more realistic...
GANimation-A study of creating facial expressions with one image
It is a code repository of GANimation, a technology that creates animations that change facial expressions by inputting a single image. Basically, it is conditional GAN, to describe the anatomical movement of the face...
Virtual Human: Saya Project
Japan's Virtual Human Project, Saya Project. It's in Japanese, so I couldn't understand all the progress, but the visual completion was quite high and the expression was natural. after…
Danbooru 2019-Animated Character Image Data
Introducing the Danbooru 2019 version link, an animated character image database. There are about 3.7 million images and about 29 tags are attached per image. Tag's…
Apple's ultra-high resolution VR headset (iGlass?)
Assuming that the human-like AI-equipped humanoid character has improved tremendously, display it on a 2D plane such as a computer or smartphone screen, and use a mouse, keyboard, and touch...
RAVDESS-Multimodal Sentiment Data
There are many complex human emotion perceptions and expressions (e.g. angry emotions affect facial expressions, voices, and language) while audio-videos are tied together...
Neural network technology by human memory characteristics
I recently read about the relationship between human abilities and neural networks. Although it appears in the article, the way the human brain and neural network work is similar, but the same...
Human-Like Testing for Candy Crush Saga
Candy Crush Saga from the famous gaming company King is a puzzle game with tons of levels. It's 2018 data, but it adds about 15 levels every week...
Replika: Emotional Chatbot
The main task of AI chatbots is to answer questions such as explaining product information, telling schedules, and checking the weather. Perhaps these…
Future of Synthetic Media (Synthesia)
This is an article from Synthesia blog that applies AI technology to media marketing with the wording “Synthetic media”. The main field of this company is the face of the model in the video…
Rosebud.AI's virtual model synthesis technology
Rosebud.ai (https://rosebud.ai/) is targeting the marketing market with a technology that creates and synthesizes virtual model faces on images created for marketing campaigns. The result is quite natural,...
How Roblox Optimizes Bert
Most chatbot systems still operate based on rules, but in order to implement natural conversations, you will eventually need to use more complex language models such as BERT…
TikTok's Comic Filter
TikTok added a filter that converts human faces into animated characters in real time. Selfie2Anime and UGatIT made by Kim Joon-ho have results for reference, but TikTok's…
