type
status
date
slug
summary
tags
category
icon
password
Last edited time
Nov 7, 2024 01:21 PM
😀

📝 主旨内容

开幕式

notion image
notion image

报告

Exploring the New Frontiers of AI – ByteDance Research's Exploration

notion image
notion image
字节跳动AI实验室:
  1. Robotic
  1. AI for Science
  1. Responsible AI
  1. AI Foundation: Large AI Models
 
蛋白质建模与设计 —— CryoFM,DPLM,DPLM2
💡
DPLM-2是一种多模态蛋白模型,通过联合序列和结构生成,提高了蛋白质建模效率和精度
dplm
bytedanceUpdated Nov 11, 2024
 
 
机器人 —— GR-1,GR-2
💡
GR-2通过视频生成预训练和机器人数据微调,实现多视角条件下的视觉操控
GR-1
bytedanceUpdated Nov 11, 2024
notion image
 
端到端同声传译 —— CLASI(Cross Language Agent – Simultaneous Interpretation)
💡
通过处理当前音频输入,结合外部知识检索和历史上下文信息,实时生成高质量的翻译。
notion image
notion image
视频生成 —— PixelDance
notion image
notion image
notion image
 
 

细粒度多模态场景理解与生成

notion image
 
 

基于大模型的神经符号计算

notion image
 

大模型检索增强

notion image
notion image
notion image
报告内容
  • 通用文本表征特征
  • 学习索引
  • RAG
notion image
notion image
💡
FlexRAG通过压缩上下文嵌入,提升生成质量并降低成本,实现灵活高效的RAG系统
 
 

海报

Vision-Language Dual-Pattern Matching for Out-of-Distribution Detection

💡
提出MCM方法,将OOD检测从单模态扩展到多模态,显著提升检测性能
ECCV2024: Vision-Language Dual-Pattern Matching for Out-of-Distribution Detection
Companion talk of ECCV2024 paper: Zihan Zhang*, Zhuo Xu*, Xiang Xiang* Vision-Language Dual-Pattern Matching for Out-of-Distribution Detection. In ECCV, 2024. Abstract: Out-of-distribution (OOD) detection is a significant challenge in deploying pattern recognition and machine learning models, as models often fail on data from novel distributions. Recent vision-language models (VLMs) such as CLIP have shown promise in OOD detection through their generalizable multimodal representations. Existing CLIP-based OOD detection methods only utilize a single modality of in-distribution (ID) information (\eg, textual cues). However, we find that the ID visual information helps to leverage CLIP's full potential for OOD detection. In this paper, we pursue a different approach and explore the regime to leverage both the visual and textual ID information. Specifically, we propose Dual-Pattern Matching (DPM), efficiently adapting CLIP for OOD detection by leveraging both textual and visual ID patterns. DPM stores ID class-wise text features as the textual pattern and the aggregated ID visual information as the visual pattern. At test time, the similarity to both patterns is computed to detect OOD inputs. We further extend DPM with lightweight adaptation for enhanced OOD detection. Experiments demonstrate DPM's advantages, outperforming existing methods on common benchmarks. The dual-pattern approach provides a simple yet effective way to exploit multi-modality for OOD detection with vision-language representations.
ECCV2024: Vision-Language Dual-Pattern Matching for Out-of-Distribution Detection
 
 
 

顶会回顾

 

ICLR——北京大学 袁粒

拒稿
转投CVPR oral
notion image
notion image
 
 

ECCV2024——重庆大学

3D视觉
 
复杂分割 华为火花奖
 
 

🤗 总结归纳

 

📎 参考文章

机场非合作目标跑道入侵检测demoMPPC——多重配对像素一致性解决瑕疵缺陷检测
Loading...