Chinese (Simplified)EnglishFrenchGermanItalianJapanesePortugueseRussianSpanish
No Result
View All Result
FuturoProssimo
Friday, April 16, 2021
  • Tech
  • Medicine
  • Society
  • Ambience
  • Spazio
  • Transportation
  • concepts
  • H+
Contact
FuturoProssimo
No Result
View All Result

Read in:
Chinese (Simplified)EnglishFrenchGermanItalianJapanesePortugueseRussianSpanish

November 8, 2019

Here's an AI that can predict the next 25 frames of a movie

Gianluca Ricciodi Gianluca Riccio
in Technology
Send to FacebookPin on PinterestSend on TwitterSend on Whatsappon Linkedin
Here's an AI that can predict the next 25 frames of a movie

An artificial intelligence is able to predict quite accurately what the 25 future frames of this movie will be.

AI and machine learning algorithms are getting better at predicting actions in videos.

The best of current algorithms can predict quite precisely where a baseball will go after it has been thrown, or the appearance of a road in the sequence to come. In other words? Predicting frames in the future of a movie.

A new approach proposed by researchers at Google, University of Michigan and Adobe advances the state of the art with large-scale models that generate high-quality video from just a few frames.

Maybe you are also interested

If AI starts prescribing drugs: the future GPT-3 doctor

The amazing FPV drones do strike in this absurd video

OncoHost, artificial intelligence that can “design” oncological therapies

2020 for neuroscience, longevity and AI, is the year to come

“With this project we aim to obtain accurate video forecasts. We will optimize the capabilities of a neural network ", the researchers wrote in a document which describes their work.

The team model

The team's core model is based on a stochastic video generation architecture, with a component that manages the predictions of the frames following those considered.

The team trained and tested different versions of the model separately from custom datasets based on three forecast categories: interactions between objects, structured movement and partial observability.

For the first task (interactions with objects) the researchers selected 256 clips from a block of videos showing a robotic arm while interacting with towels.

For the second (structured movement) they edited clips from Human 3.6M, a block containing clips of humans performing actions like sitting in a chair.

As for the third (partial observability activity), used an open source KITTI driving dataset collected from footage of cameras mounted on car dashboards.

After this "training," the AI ​​model generated up to 25 frames in the future.

Researchers report that "predictions" were preferred 90,2, 98,7% and 99,3% of the time by evaluators respectively to the three types of video: interactions between objects, structured movement and partial observability tasks. respectively.

Qualitatively, the team notes that the AI ​​neatly represented human arms and legs and done "Very precise predictions that seemed realistic compared to the scenes represented in the video" .

The artificial intelligence model while providing video data frames of human actions
The artificial intelligence model while providing video data frames of a car camera.

"We found that maximizing the capacity of such models improves the quality of the video prediction," coauthors write. We hope our work will encourage the field to move in similar directions in the future. For example to see how far we can go ".

How many minutes of the future can you imagine?

tags: AIForecastsVideo
Previous post

OpenAI makes public AI capable of writing completely credible fake news

Next article

Bill Gates finances an unsinkable metal

Collaborate!

We are open to visions about the future. Submit an article, disclose the results of a search or scientific discoveries, shows points of view on a theme, tells about a change.

Contact us

Most read of the week

  • US intelligence analyzes the future, and it doesn't look good at all

    US intelligence analyzes the future, and it doesn't look good at all

    96 shares
    Share 38 Tweet 24
  • Physicists at Microsoft think the universe is a self-taught computer

    66 shares
    Share 26 Tweet 16
  • Plants that “exude” metal: can agromining replace mines?

    65 shares
    Share 26 Tweet 16
  • Earth 300, the crazy gigayacht surmounted by a giant sphere

    61 shares
    Share 24 Tweet 15
The last
Architecture

Parco Romana, in Milan a station becomes a tree-lined park

Ambience

An autonomous SmartFarm grows plants with water vapor taken from the air

Ambience

The whitest white paint there is is even whiter (and cools)

Spazio

5 star systems capable of hosting life have just been discovered

archive

Next article
Bill Gates finances an unsinkable metal

Bill Gates finances an unsinkable metal

Penis transplant

Doctors announce the success of the first, er, penis transplant

Facebook

Instagram

Telegram

Twitter

Clubhouse

Near future

Futuroprossimo.it is an Italian futurology resource open since 2006: every day news about the near future. Scientific discoveries, medical research, prototypes, concepts and predictions about the future for free.

Tag

Ambience Architecture Club Communication concepts Advice Economy Energy Events Gadgets The future of yesterday The newspaper of tomorrow Italy Next Medicine Military Weather Robotica Society Spazio Technology transhumanism Transportation Video

The author

Gianluca Riccio, copywriter and journalist - Born in 1975, he is the creative director of an advertising agency, he is affiliated with the Italian Institute for the Future, World Future Society and H +, Network of Italian Transhumanists.

Home / Author / IDEA / archive / Promo on FP

Collaborate! Are you interested in writing a post on Futuroprossimo? Click here for contacts.

Categories

Creative Commons License
This work is distributed under license Creative Commons Attribution 4.0 International.
© 2021 Futuroprossimo

  • Home
  • Contact
  • archive
  • Technology
  • Medicine
  • Transportation
  • Weather
  • Society
  • Ambience
  • transhumanism

© 2021 Futuroprossimo - Creative Commons License
This work is distributed under license Creative Commons Attribution 4.0 International.

This site uses cookies. By continuing to read it, you consent to their use.