Close Menu
Ztoog
    What's Hot
    The Future

    Razer launches new gaming range at CES

    Gadgets

    Make waves in 2025: Exhibit at Ztoog events

    Mobile

    OpenAI’s official ChatGPT app for Android arrives on the Play Store

    Important Pages:
    • About Us
    • Contact us
    • Privacy Policy
    • Terms & Conditions
    Facebook X (Twitter) Instagram Pinterest
    Facebook X (Twitter) Instagram Pinterest
    Ztoog
    • Home
    • The Future

      How to Get Bot Lobbies in Fortnite? (2025 Guide)

      Can work-life balance tracking improve well-being?

      Any wall can be turned into a camera to see around corners

      JD Vance and President Trump’s Sons Hype Bitcoin at Las Vegas Conference

      AI may already be shrinking entry-level jobs in tech, new research suggests

    • Technology

      What does a millennial midlife crisis look like?

      Elon Musk tries to stick to spaceships

      A Replit employee details a critical security flaw in web apps created using AI-powered app builder Lovable that exposes API keys and personal info of app users (Reed Albergotti/Semafor)

      Gemini in Google Drive can now help you skip watching that painfully long Zoom meeting

      Apple iPhone exports from China to the US fall 76% as India output surges

    • Gadgets

      Watch Apple’s WWDC 2025 keynote right here

      Future-proof your career by mastering AI skills for just $20

      8 Best Vegan Meal Delivery Services and Kits (2025), Tested and Reviewed

      Google Home is getting deeper Gemini integration and a new widget

      Google Announces AI Ultra Subscription Plan With Premium Features

    • Mobile

      YouTube is testing a leaderboard to show off top live stream fans

      Deals: the Galaxy S25 series comes with a free tablet, Google Pixels heavily discounted

      Microsoft is done being subtle – this new tool screams “upgrade now”

      Wallpaper Wednesday: Android wallpapers 2025-05-28

      Google can make smart glasses accessible with Warby Parker, Gentle Monster deals

    • Science

      Some parts of Trump’s proposed budget for NASA are literally draconian

      June skygazing: A strawberry moon, the summer solstice… and Asteroid Day!

      Analysts Say Trump Trade Wars Would Harm the Entire US Energy Sector, From Oil to Solar

      Do we have free will? Quantum experiments may soon reveal the answer

      Was Planet Nine exiled from the solar system as a baby?

    • AI

      Fueling seamless AI at scale

      Rationale engineering generates a compact new tool for gene therapy | Ztoog

      The AI Hype Index: College students are hooked on ChatGPT

      Learning how to predict rare kinds of failures | Ztoog

      Anthropic’s new hybrid AI model can work on tasks autonomously for hours at a time

    • Crypto

      Bitcoin Maxi Isn’t Buying Hype Around New Crypto Holding Firms

      GameStop bought $500 million of bitcoin

      CoinW Teams Up with Superteam Europe to Conclude Solana Hackathon and Accelerate Web3 Innovation in Europe

      Ethereum Net Flows Turn Negative As Bulls Push For $3,500

      Bitcoin’s Power Compared To Nuclear Reactor By Brazilian Business Leader

    Ztoog
    Home » UC Berkeley and Microsoft Research Redefine Visual Understanding: How Scaling on Scales Outperforms Larger Models with Efficiency and Elegance
    AI

    UC Berkeley and Microsoft Research Redefine Visual Understanding: How Scaling on Scales Outperforms Larger Models with Efficiency and Elegance

    Facebook Twitter Pinterest WhatsApp
    UC Berkeley and Microsoft Research Redefine Visual Understanding: How Scaling on Scales Outperforms Larger Models with Efficiency and Elegance
    Share
    Facebook Twitter LinkedIn Pinterest WhatsApp

    In the dynamic realm of laptop imaginative and prescient and synthetic intelligence, a brand new method challenges the normal pattern of constructing bigger fashions for superior visible understanding. The method within the present analysis, underpinned by the assumption that bigger fashions yield extra highly effective representations, has led to the event of gigantic imaginative and prescient fashions. 

    Central to this exploration lies a important examination of the prevailing follow of mannequin upscaling. This scrutiny brings to gentle the numerous useful resource expenditure and the diminishing returns on efficiency enhancements related with constantly enlarging mannequin architectures. It raises a pertinent query in regards to the sustainability and effectivity of this method, particularly in a website the place computational assets are invaluable.

    UC Berkeley and Microsoft Research launched an modern method referred to as Scaling on Scales (S2). This technique represents a paradigm shift, proposing a method that diverges from the normal mannequin scaling. By making use of a pre-trained, smaller imaginative and prescient mannequin throughout numerous picture scales, S2 goals to extract multi-scale representations, providing a brand new lens via which visible understanding will be enhanced with out essentially rising the mannequin’s dimension.

    Leveraging a number of picture scales produces a composite illustration that rivals or surpasses the output of a lot bigger fashions. The analysis showcases the S2 method’s prowess throughout a number of benchmarks, the place it constantly outperforms its bigger counterparts in duties together with however not restricted to classification, semantic segmentation, and depth estimation. It units a brand new state-of-the-art in multimodal LLM (MLLM) visible element understanding on the V* benchmark, outstripping even industrial fashions like Gemini Pro and GPT-4V, with considerably fewer parameters and comparable or diminished computational calls for.

    For occasion, in robotic manipulation duties, the S2 scaling technique on a base-size mannequin improved the success charge by about 20%, demonstrating its superiority over mere model-size scaling. The detailed understanding functionality of LLaVA-1.5, with S2 scaling, achieved outstanding accuracies, with V* Attention and V* Spatial scoring 76.3% and 63.2%, respectively. These figures underscore the effectiveness of S2 and spotlight its effectivity and the potential for decreasing computational useful resource expenditure.

    This analysis sheds gentle on the more and more pertinent query of whether or not the relentless scaling of mannequin sizes is really vital for advancing visible understanding. Through the lens of the S2 method, it turns into evident that various scaling strategies, notably these focusing on exploiting the multi-scale nature of visible information, can present equally compelling, if not superior, efficiency outcomes. This method challenges the present paradigm and opens up new avenues for resource-efficient and scalable mannequin growth in laptop imaginative and prescient.

    In conclusion, introducing and validating the Scaling on Scales (S2) technique represents a big breakthrough in laptop imaginative and prescient and synthetic intelligence. This analysis compellingly argues for a departure from the prevalent mannequin dimension growth in the direction of a extra nuanced and environment friendly scaling technique that leverages multi-scale picture representations. Doing so demonstrates the potential for attaining state-of-the-art efficiency throughout visible duties. It underscores the significance of modern scaling methods in selling computational effectivity and useful resource sustainability in AI growth. The S2 technique, with its skill to rival and even surpass the output of a lot bigger fashions, affords a promising various to conventional mannequin scaling, highlighting its potential to revolutionize the sphere.


    Check out the Paper and Github. All credit score for this analysis goes to the researchers of this mission. Also, don’t overlook to observe us on Twitter. Join our Telegram Channel, Discord Channel, and LinkedIn Group.

    If you want our work, you’ll love our publication..

    Don’t Forget to affix our 39k+ ML SubReddit


    Sana Hassan, a consulting intern at Marktechpost and dual-degree pupil at IIT Madras, is keen about making use of know-how and AI to handle real-world challenges. With a eager curiosity in fixing sensible issues, he brings a recent perspective to the intersection of AI and real-life options.


    🐝 Join the Fastest Growing AI Research Newsletter Read by Researchers from Google + NVIDIA + Meta + Stanford + MIT + Microsoft and many others…

    Share. Facebook Twitter Pinterest LinkedIn WhatsApp

    Related Posts

    AI

    Fueling seamless AI at scale

    AI

    Rationale engineering generates a compact new tool for gene therapy | Ztoog

    AI

    The AI Hype Index: College students are hooked on ChatGPT

    AI

    Learning how to predict rare kinds of failures | Ztoog

    AI

    Anthropic’s new hybrid AI model can work on tasks autonomously for hours at a time

    AI

    AI learns how vision and sound are connected, without human intervention | Ztoog

    AI

    How AI is introducing errors into courtrooms

    AI

    With AI, researchers predict the location of virtually any protein within a human cell | Ztoog

    Leave A Reply Cancel Reply

    Follow Us
    • Facebook
    • Twitter
    • Pinterest
    • Instagram
    Top Posts
    The Future

    How it’s Changing the Digital Marketing Landscape

    In the age of good audio system and voice-activated assistants, the means customers seek for…

    Science

    Starship launch 3: What time is the SpaceX flight and what to expect?

    SpaceX’s Starship prepped for flightSpaceX SpaceX is launching its huge Starship rocket for the third…

    The Future

    Top 7 employee productivity tracking software

    To save time and enhance effectivity throughout your crew, you may need tried implementing to-do…

    The Future

    Microsoft Surface Laptop Studio 2 and Laptop Go 3: how to preorder

    Although the forthcoming laptops look comparable in design to their predecessors, the brand new Surface…

    Science

    Quantum entanglement can be endlessly ’embezzled’ from quantum fields

    Some quantum fields have an infinite quantity of entanglement that can be takenJurik Peter/Shutterstock Entanglement…

    Our Picks
    Crypto

    Why Is Bitcoin Price Down Today?

    Mobile

    Black Friday: Get a great deal on Galaxy Watch6, iPad 10.9″ 2022 and more in Germany

    The Future

    House Oversight Committee Will Hold Hearing on UFOs

    Categories
    • AI (1,494)
    • Crypto (1,754)
    • Gadgets (1,806)
    • Mobile (1,852)
    • Science (1,868)
    • Technology (1,804)
    • The Future (1,650)
    Most Popular
    Crypto

    NFT startup Rario founders to leave a year after $120M funding

    Gadgets

    OPPO Air Glass 3 Unveiled Bringing AI And Voice Assistant

    Gadgets

    Google’s new “inactive account” policy won’t delete years of YouTube videos

    Ztoog
    Facebook X (Twitter) Instagram Pinterest
    • Home
    • About Us
    • Contact us
    • Privacy Policy
    • Terms & Conditions
    © 2025 Ztoog.

    Type above and press Enter to search. Press Esc to cancel.