Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[TRN] Temporal Relational Reasoning in Videos #2

Closed
jinsoo9595 opened this issue Jul 28, 2022 · 1 comment
Closed

[TRN] Temporal Relational Reasoning in Videos #2

jinsoo9595 opened this issue Jul 28, 2022 · 1 comment
Assignees

Comments

@jinsoo9595
Copy link
Owner

  • 당시 문제점

    • Motion 특징 추출을 optical flow에 의존함 → 시스템 효율을 낮춤
    • 3D Conv는 dense frame을 처리하기 때문에 computation이 많이 요구됨
  • 해결 방안 및 특징

    • Multi-scale temporal input
    • 2D CNN 기반 모델에 쉽게 적용 가능
      • Two-stream(spatial & temporal) 모델에 적용할 경우 성능 더욱 향상
  • 동작 과정
    Image

    • Input
      • Video clip에서 2, 3, …, N frames를 시간순으로 입력 (𝑁=2~8)
      • 𝐾(=3) relation만 확인 (2/3/4-frames, 3/4/5-frames, …)
    • BN-Inception 모델 이용
    • Feature map → 2-layer MLP(unit: 256) → 1-layer MLP(unit: class number)
    • Global average pooling layer
      Image
@jinsoo9595 jinsoo9595 self-assigned this Jul 28, 2022
@jinsoo9595
Copy link
Owner Author

.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
Development

No branches or pull requests

1 participant