Popular Posts

Smilegate.AI's homepage has been renewed and newly opened. In addition to providing technologies and data that fit the direction of the AI strategy pursued by Smilegate according to human functions, it will be used for the purpose of disclosing the results of future projects at Smilegate.AI. Each technology is visual, hearing, and feet corresponding to human appearance...

■ Smilegate Group establishes an AI center specialized in the entertainment industry ■ Fun AI that pursues fun and excitement rather than providing simple information, Human-like AI development goal that pursues comprehensive humanization technology rather than realizing human unit functions [2020-0803] Smilegate Is an AI specialized in the entertainment industry based on artificial intelligence (AI) technology...

Most chatbot systems still operate based on rules, but in order to implement natural conversations, you eventually need to use more complex language models such as BERT. However, there is a lot of recognition that BERT is heavy and complex, and we briefly introduced what the game platform company Roblox did to make it a service using BERT. Basically, servicing using a GPU...

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. Apart from the "software capabilities" that the human brain shows, I wondered how much should I increase to reach the "hardware" capabilities of the human brain? (If you reach the same hardware capabilities, you can now see the algorithmic efficiency of human vs. AI...

We are looking for active talents who have the ability to develop new ideas in various fields and fast prototyping new ideas. The main task is to efficiently develop services that can create practical value using various AI technologies. Those who have an overview of the AI field and experience participating in development projects are preferred. ...

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. The robot cries at constant demand and eventually breaks it down. Watching the video, my heart hurt a lot. The robot that expresses the mind and interacts with it moves the human mind. The idea that this is a human-like robot...

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. Recruitment will take place in various fields ranging from innovative technology research, optimization for service application, and further creation of new services. In addition, open source contributors, open community influencers, and others who have conducted open research...

The main task of AI chatbots is to answer questions such as explaining product information, telling schedules, and checking the weather. Perhaps if these technologies are maximized, we can reach'Super Human AI' for some of the human fields of'knowledge'. However, with "Human-Like AI", you need to worry about the overall human ability rather than maximizing specific functions...

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 technology seems to have reached a considerable level in terms of completeness. Live2D, which requires a separate marking of delicate face parts, FaceRi, which requires a pre-made 3D model...

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. It is said that in the future, it is planned to add functions such as adding human emotions, behavior recognition, and dialogue... In order to create a human AI, various things such as appearance, voice, dialogue, and emotions ...

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 technology seems to have reached a considerable level in terms of completeness. Live2D, which requires a separate marking of delicate face parts, FaceRi, which requires a pre-made 3D model...

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. It is said that in the future, it is planned to add functions such as adding human emotions, behavior recognition, and dialogue... In order to create a human AI, various things such as appearance, voice, dialogue, and emotions ...

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, and it has the advantage of being able to replace the model face at any time without having to take a picture again. The main advantage is that you can save expensive model licensing fees...

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 is called MotionScan. Basically, the cost is very high because the actor sits in a room where several cameras are elaborately arranged to perform, capture and reprocess it. So, a large game company...

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. Examples of tags include "1girl", "solo", "long_hair", "highres", "smile", and "open_mouth", which can be used for various purposes. ring...

There are many complex human emotion perceptions and expressions (e.g. angry emotions affect facial expressions, voices, and language). Here is an open dataset with audio-videos tied together and emotionally labeled. It's called The Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS)...

It is a code repository of GANimation, a technology that creates animations that change facial expressions by inputting a single image. Basically, it is a conditional GAN, and it uses FACS (facial action coding system), a methodology to describe the anatomical movements of the face. According to FACS, our face utilizes approximately 43 muscles for a total...

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 appearance has been changed to be more realistic and shows the technique of tracking the entire body without a marker. When the Codec Avatar was first released in March 2019, the human head and face were elaborately modeled...

This is an article from Synthesia blog that applies AI technology to media marketing with the wording "Synthetic media". The company's main field is to move the facial expressions and lip shape of the model in the video from another video and synthesize (adjust?) it. For example, when there is a video taken by Beckham (speaking in English), the video in 9 different languages...

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 are uploaded, and you can download them in a 3D format called (Autodesk) FBX. This format can be read in 3ds or Maya, and can be used in Unity or Unreal Engine. Character motion is quite special...

There are many complex human emotion perceptions and expressions (e.g. angry emotions affect facial expressions, voices, and language). Here is an open dataset with audio-videos tied together and emotionally labeled. It's called The Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS)...

After performing representation training with 53,000 hours of label-free data, a pre-trained model for Facebook's wav2vec 2.0, which became a hot topic because it created a speech recognizer with only 10 minutes of labeled data, was released. Representation model no fine-tuning, 10 minutes,...

I've heard stories that you can know what you're talking about with just the movement of your lips if you receive special training, but the research in the link is realized with AI. , The attention-based speech decoder generates mel cepstrum. After that...

TensorflowTTS, an open source based on Tensorflow 2 that supports several latest TTS models such as Tacotron2, MelGan, FastSpeech, etc., has finally started supporting Microsoft FastSpeech2. FastSpeech2 shows similar performance to Transformer series TTS, but the training time has been reduced by more than 2 times and...

MIT's Speech2Face is a study that generates a speaker's face from a speech signal. However, it does not perform speech to face transform with one model, and it combines the results of existing studies for different purposes to create impressive results. (The first author is Professor Tae-Hyun Oh currently at Pohang University) The first is facial vect...

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 was somewhat disappointing in terms of visual artifacts and the naturalness of movement. Consider temporal correlation using frame, simple...

Corona 19 has yet to show signs of calming down worldwide. At MIT, we learned an AI model that can check whether COVID-19 is infected from the cough sound recorded with a mobile phone, and published the methodology and experimental results for this as a thesis. According to the experimental results, not only was it possible to predict the presence of Corona 19 infection with an accuracy of about 98%, but even asymptomatic infection...

Most chatbot systems still operate based on rules, but in order to implement natural conversations, you eventually need to use more complex language models such as BERT. However, there is a lot of recognition that BERT is heavy and complex, and we briefly introduced what the game platform company Roblox did to make it a service using BERT. Basically, servicing using a GPU...

The main task of AI chatbots is to answer questions such as explaining product information, telling schedules, and checking the weather. Perhaps if these technologies are maximized, we can reach'Super Human AI' for some of the human fields of'knowledge'. However, with "Human-Like AI", you need to worry about the overall human ability rather than maximizing specific functions...

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. The robot cries at constant demand and eventually breaks it down. Watching the video, my heart hurt a lot. The robot that expresses the mind and interacts with it moves the human mind. The idea that this is a human-like robot...

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 networks operate have similar corners, but I don't think they need to be the same. Rather than imitating fish to swim, you can make a submarine, and make an airplane rather than imitate a bird to fly. (This is the author's example).

Candy Crush Saga from the famous gaming company King is a puzzle game with tons of levels. It's 2018 data, but it's said to add about 15 levels every week. (Approximately 3700 per year) I need to measure and balance the level of difficulty, but the problem is that it takes one week per level for human testers. To solve this problem, a human-like way...

There are many complex human emotion perceptions and expressions (e.g. angry emotions affect facial expressions, voices, and language). Here is an open dataset with audio-videos tied together and emotionally labeled. It's called The Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS)...

Many MRC models proposed so far show evaluation values beyond human capabilities in various tasks and datasets, but I think it is difficult to easily say YES when asked if you understand a given context better than humans? First of all, there are various questions about whether the current MRC techniques really "understand"...

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. In the world, a lot of data is built as a relational database, and the standard language used to acquire information from this database is S...

Scatterlab (https://scatterlab.co.kr/), which stands out in everyday conversational research, is an article on the Ping-Pong team blog. I still see GPT-3 as a'eye of suspicion', but it must be strange to look at it again.

The performance improvement that Transformer-based language models show is surprising, but as the model size increases exponentially, concerns about service costs are also becoming important. In the case of Bert-base or GPT-2, there are about 100 million parameters, so the model size, memory bandwidth, and inference time all have to be optimized to a significant degree. Vs used for model optimization...

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. Apart from the "software capabilities" that the human brain shows, I wondered how much should I increase to reach the "hardware" capabilities of the human brain? (If you reach the same hardware capabilities, you can now see the algorithmic efficiency of human vs. AI...

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. The robot cries at constant demand and eventually breaks it down. Watching the video, my heart hurt a lot. The robot that expresses the mind and interacts with it moves the human mind. The idea that this is a human-like robot...

It is said that the learning cost of GPT-3, the pronoun of the super-scale language model, which shows the possibility that it can be applied to all natural language tasks only with Few shot learning, is estimated at 4 billion KRW in Korean. About 40 million won for GPT-2. No matter how highly recyclable it is, it is difficult for most companies to invest in R&D. (Someone created and distributed KoGPT-3...

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. Rather than borrowing a book from a library and "searching" for information as you read it sequentially, you use a search engine to find the information you want right in a huge, already full-text indexed knowledge base. But, Cisco kid...

In the field of prediction, Algorithm Aversion is used to mean that once you realize that an algorithm can make mistakes, you tend to avoid using it, even if it is better than human predictions. For example, in the case of AI interviews, disease prediction, autonomous driving, and AI judges, even if all technical difficulties are solved and the average performance is better than humans...

Scatterlab (https://scatterlab.co.kr/), which stands out in everyday conversational research, is an article on the Ping-Pong team blog. I still see GPT-3 as a'eye of suspicion', but it must be strange to look at it again.

Assuming that human-like AI-equipped humanoid characters have improved enormously, it is not possible to use their full potential just by displaying them on a 2D plane such as a computer or smartphone screen and interacting with the mouse, keyboard, and touch. I think. The alternatives to this are technologies like AR or VR, but in terms of realism and operation...

According to various statistics, the number of AI-related major conference participants increased 6 times over 5 years, and the number of AI startups increased by 113% while non-AI startups increased 28% over 4 years. Venture investments in the non-AI field increased by 100% over 5 years, while 350% increased, and deep learning-related jobs increased by 34 times over 3 years. AI, deep learning...

Share what you've recently enjoyed reading. In this article, we will introduce a form in which humans and AI work in different fields, and one side collaborates rather than replaces the other. Collective intelligence refers to a phenomenon in which the intellectual abilities of a group are combined to exert greater abilities, not one's intellectual abilities. In the AI field, W...

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, depthwise separable convolution that applies convolution with a size of 1x1xC in the channel direction greatly reduces the number of parameters...

The main task of AI chatbots is to answer questions such as explaining product information, telling schedules, and checking the weather. Perhaps if these technologies are maximized, we can reach'Super Human AI' for some of the human fields of'knowledge'. However, with "Human-Like AI", you need to worry about the overall human ability rather than maximizing specific functions...

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 is called MotionScan. Basically, the cost is very high because the actor sits in a room where several cameras are elaborately arranged to perform, capture and reprocess it. So, a large game company...

It is a code repository of GANimation, a technology that creates animations that change facial expressions by inputting a single image. Basically, it is a conditional GAN, and it uses FACS (facial action coding system), a methodology to describe the anatomical movements of the face. According to FACS, our face utilizes approximately 43 muscles for a total...

After performing representation training with 53,000 hours of label-free data, a pre-trained model for Facebook's wav2vec 2.0, which became a hot topic because it created a speech recognizer with only 10 minutes of labeled data, was released. Representation model no fine-tuning, 10 minutes,...

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. In the world, a lot of data is built as a relational database, and the standard language used to acquire information from this database is S...

I've heard stories that you can know what you're talking about with just the movement of your lips if you receive special training, but the research in the link is realized with AI. , The attention-based speech decoder generates mel cepstrum. After that...

Scatterlab (https://scatterlab.co.kr/), which stands out in everyday conversational research, is an article on the Ping-Pong team blog. I still see GPT-3 as a'eye of suspicion', but it must be strange to look at it again.

TensorflowTTS, an open source based on Tensorflow 2 that supports several latest TTS models such as Tacotron2, MelGan, FastSpeech, etc., has finally started supporting Microsoft FastSpeech2. FastSpeech2 shows similar performance to Transformer series TTS, but the training time has been reduced by more than 2 times and...

The performance improvement that Transformer-based language models show is surprising, but as the model size increases exponentially, concerns about service costs are also becoming important. In the case of Bert-base or GPT-2, there are about 100 million parameters, so the model size, memory bandwidth, and inference time all have to be optimized to a significant degree. Vs used for model optimization...

Typically, Q&A systems use text to answer questions. A task in this way is the Squad task, which gives you a paragraph explaining a fact and asks a question to generate an appropriate answer. In contrast, Visual QA is a task of giving an image instead of text and having a conversation about it. Today’s paper, “3D Atte...

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. Examples of tags include "1girl", "solo", "long_hair", "highres", "smile", and "open_mouth", which can be used for various purposes. ring...

There are many complex human emotion perceptions and expressions (e.g. angry emotions affect facial expressions, voices, and language). Here is an open dataset with audio-videos tied together and emotionally labeled. It's called The Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS)...

Many MRC models proposed so far show evaluation values beyond human capabilities in various tasks and datasets, but I think it is difficult to easily say YES when asked if you understand a given context better than humans? First of all, there are various questions about whether the current MRC techniques really "understand"...

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. In the world, a lot of data is built as a relational database, and the standard language used to acquire information from this database is S...

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 are uploaded, and you can download them in a 3D format called (Autodesk) FBX. This format can be read in 3ds or Maya, and can be used in Unity or Unreal Engine. Character motion is quite special...

I've heard stories that you can know what you're talking about with just the movement of your lips if you receive special training, but the research in the link is realized with AI. , The attention-based speech decoder generates mel cepstrum. After that...

This 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. Label is "vehicle, pedestrian, road, sidewalk, bicycle, motorcycle, building, ter...

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, if you give a picture of a white cat and a black dog together and ask, "What color is the animal next to the cat?", you answer "black". Specifically, if you give an image, dialog history, and question, answe...

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 it is said that anyone can download and use the file once the online agreement is written and approved on the'Everyone's Corpus' site. Compared to the last '21st Century Sejong Plan', the data constructed this time is a ratio of colloquial materials such as daily conversations, messengers, web documents...

The first stable version 1.0 of the Huggingface Datasets library has been released, making it easy to use NLP datasets and evaluation metrics. Currently, about 100 datasets and evaluation metrics (about 10) for each dataset are supported. Ten...

Smilegate.AI's homepage has been renewed and newly opened. In addition to providing technologies and data that fit the direction of the AI strategy pursued by Smilegate according to human functions, it will be used for the purpose of disclosing the results of future projects at Smilegate.AI. Each technology is visual, hearing, and feet corresponding to human appearance...

■ Smilegate Group establishes an AI center specialized in the entertainment industry ■ Fun AI that pursues fun and excitement rather than providing simple information, Human-like AI development goal that pursues comprehensive humanization technology rather than realizing human unit functions [2020-0803] Smilegate Is an AI specialized in the entertainment industry based on artificial intelligence (AI) technology...

■ SeAH, streaming broadcast debut! Live broadcasting officially launched in August ■ Expect more broadcasting activities with the Sandbox Network! [2020-0730] Smilegate Entertainment (CEO Jang In-ah), Korea's first virtual YouTuber'SE:A', exclusively works with Sandbox Network, a leading digital entertainment company in the MCN industry, for full-fledged broadcasting activities...

Active communication and reaction with the contents desired by viewers on SeAH Live Broadcasting Twitch broadcast, which was launched at the same time as the start of'Good' Season 2, also increased subscribers reaching 10,000 [2020-0814] Smilegate Entertainment (CEO Jang In-ah) is Korea's first Virtual YouTuber'SE:A' started live broadcasting after a test broadcast for the past one month.

Orange Planet Jeonju Center's tenant, Medi-K System, is a healthcare company that develops oriental medicine bath room solutions based on ICT (information and communication technology). By incorporating new technologies of the 4th industrial revolution with the oriental medicine industry, it is bringing new winds to the oriental medicine world. In recent years, the demand for systematic herbal medicine preparation management is increasing in the oriental medicine industry, but there has been no systematic system to support this. Joint with Oriental Clinic...

'Creativity' is a staple item that has not been left out of the talent perspective of recent companies. In order to become a leader rather than a follower in a rapidly changing era, creativity to solve problems with ideas that have never been before is essential. However, it is very difficult to express creativity alone, in a rigid environment. An environment in which you can dream freely and present ideas, and a relationship in which you can grow while thinking together...

AI-based real-time text interpretation software service support for children and adolescents with hearing impairments in the blind spot in the online education environment caused by COVID-19 MAP Tool Pack Crowdfunding Proceeds and Future Lab Workshop Participants' Donations… Expected to spread the culture of'Virtual Cycle of Sharing' [2020-1021] Creative Environment Research Center Smilegate Future Lab (hereinafter referred to as Future Lab) developed its own creative learning ...

Smilegate.AI's facebook page has been opened. On the homepage of this place, we will mainly share technical and professional content, while on the facebook page, we will introduce the center news, research process, and small talks on life. For example, a post recently posted is about the day of education for new hires^^ Today is a depressing day.'New ...

We are looking for active talents who have the ability to develop new ideas in various fields and fast prototyping new ideas. The main task is to efficiently develop services that can create practical value using various AI technologies. Those who have an overview of the AI field and experience participating in development projects are preferred. ...

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. Recruitment will take place in various fields ranging from innovative technology research, optimization for service application, and further creation of new services. In addition, open source contributors, open community influencers, and others who have conducted open research...