namul2story

  • 홈
  • 태그
  • 방명록

AI 2

[논문리뷰] VQ-VAE: Neural Discrete Representation Learning

해당 논문은 2018년 4월, Deepmind에서 나온 논문으로 Vector Quantised Variational AutoEncoder, VQ-VAE 라고 부르는 모델을 소개한다. 기존 VAE와 Vector quantisation을 접목해 나온 method라고 생각하면 될것 같다. 논문에서 스펠링 보고 헷갈렸는데 찾아보니 영국식 영어로 learnt, quantised를 사용하고 있다.Variational AutoEncoder (VAE)기존 VAE같은 경우 위와 같은 구조를 갖고있는데 input을 encoder를 통해 input x 에대해 latent vector z의 분포 (mean, std)를 예측하는 형태이다. 즉 encoder는 q(z|x)를 approximate하는 것을 목표로 한다. decod..

AI/Representation Learning 2024.09.18

[논문리뷰] SAM 2 : Segment Anything in Image and Videos

SAM 2 : Segment Anything in Image and VideosSAM2은 2024년 7월 meta ai research에서 공개한 모델로 Segment Anything (5, Apr 2023)의 이미지에서 동영상까지 처리할 수 있는 진화버전이다. SAM (Segment Anything model)SAM2를 정리하기 전에 기존 SAM은 어떤 형태를 가지고 있는지 보려고한다.SAM paper 논문에서 목표는 building a foundation model for image segmentation 이라고 한다. (foundation model: models built on unlabeled data using self-supervision, 다양한 task에 대해서 downstream 작업을..

AI/Computer Vision 2024.09.18
이전
1
다음
더보기
프로필사진

namul2story

머리굴리는 블로그

  • 분류 전체보기 (10)
    • Robotics (2)
      • Imitation Learning (1)
      • Reinforcement Learning (0)
    • AI (0)
      • Computer Vision (1)
      • Natural Language Process & .. (0)
      • Representation Learning (1)
      • Reinforcement Learning (0)
    • Project (0)
    • 뻘글 (4)
    • 용어 정리 (1)

Tag

AI, imitation learning, paper review, Computer Vision, representation, robotics, 3D,

최근글과 인기글

  • 최근글
  • 인기글

최근댓글

공지사항

페이스북 트위터 플러그인

  • Facebook
  • Twitter

Archives

Calendar

«   2025/05   »
일 월 화 수 목 금 토
1 2 3
4 5 6 7 8 9 10
11 12 13 14 15 16 17
18 19 20 21 22 23 24
25 26 27 28 29 30 31

방문자수Total

  • Today :
  • Yesterday :

Copyright © Kakao Corp. All rights reserved.

티스토리툴바