Search AI Products and News

Voice assistants are increasingly integral to our lives, yet existing digital assistants often feel flat and lack emotional and human elements in their interactions. Sesame is tackling this problem, striving for a novel "voice presence" that allows digital assistants to be more authentic, understood, and valued in communication. Sesame's core goal is to create a digital companion, not just a tool for processing requests, but a partner capable of genuine conversation. These digital companions...

Driven by multimodal large language models (MLLMs), significant advancements have been made in tasks related to images and videos, including visual question answering, narrative generation, and interactive editing. However, achieving fine-grained understanding of video content still poses major challenges. These challenges involve tasks such as pixel-level segmentation, tracking with language descriptions, and visual question answering based on specific video prompts. Although current state-of-the-art video perception models excel in segmentation and tracking tasks, they still fall short in open language understanding and conversational capabilities.

In the field of video analysis, the persistence of objects is an important cue for humans to understand that objects still exist even when completely occluded. However, current object segmentation methods mainly focus on visible (modal) objects, lacking the capability to handle non-modal (visible + invisible) objects. To address this issue, researchers proposed a two-stage method based on diffusion priors, Diffusion-Vas, aimed at improving the effectiveness of non-modal segmentation and content completion in videos, enabling the tracking of specified targets within the video and using diffusion models to complete the occluded parts.

A research team composed of Renmin University of China, Beijing University of Posts and Telecommunications, and Shanghai AI Lab has developed a technology called Ref-AVS, aimed at solving the challenge of artificial intelligence understanding the complex physical world. This technology utilizes a unique multimodal fusion method that integrates Video Object Segmentation (VOS), Video Object Reference Segmentation (Ref-VOS), and audiovisual segmentation (AVS) information, enabling AI systems to accurately identify and locate specific objects in audiovisual scenes, regardless of whether those objects produce sound. To validate the effectiveness of the technology, the research team...

MedSAM-2, based on the Segment Anything Model 2 (SAM2) framework, brings revolutionary changes to the field of medical image segmentation. Its core breakthrough treats 2D and 3D images as video sequences, enabling efficient segmentation of three-dimensional images. The 'one-click segmentation' feature allows users to specify a target in a single image, and the model can automatically recognize and accurately segment the same type of object in subsequent images without the need for continuous input. As a general model, it can handle any image.

At this week's SIGGRAPH conference, Meta CEO Mark Zuckerberg showcased Segment Anything 2 (SA2) for the first time, an upgraded version of the groundbreaking image segmentation model the company introduced last year. The new model extends AI-driven segmentation technology to the video domain, demonstrating the remarkable progress made over the past year.

RoboflowSports is a deep learning-based sports data analysis tool that revolutionizes the field with advanced object detection, image segmentation, and keypoint detection. It accurately captures game details, such as ball trajectories and athlete movements, providing coaches and analysts with deep insights. The tool's visualization features present data intuitively and support dataset management, aiding tactical research, performance evaluation, ....

Recently, several artificial intelligence large models have garnered widespread attention for making errors in simple numerical comparisons. Prominent AI models, including ByteBean, GPT4o, Kimi from the Dark Side of the Moon, StepStar JumpAsk, and Baichuan Intelligence's BaiXiaoYing, all provided incorrect answers to basic questions like "Which is larger, 9.11 or 9.9?" Additionally, earlier reports indicated that multiple large models incorrectly answered how many "r"s are in the word "strawberr

Bilibili made its debut at the 2024 World Artificial Intelligence Conference (WAIC2024), showcasing its latest self-developed AI technologies and new trends in AI-generated content. The exhibition included technologies for quickly creating digital avatars, dynamic manga techniques, and a custom AI voice library for the first Chinese virtual singer, Luo Tianyi, which drew considerable attention from the audience. As one of the most influential internet platforms in China in AI intelligence, Bilib

AI large models + multi-scale object-oriented segmentation solve the issue of object segmentation; multiple AI models + object-oriented extraction methods address the problem of object extraction; multi-scale object-oriented editing + vector smoothing export resolves issues with object correction and mapping.