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

Deep Image Aesthetics Classification using Inception Modules and Fine-tuning Connected Layer #99

Open
chullhwan-song opened this issue Feb 22, 2019 · 1 comment

Comments

@chullhwan-song
Copy link
Owner

http://jinxin.me/downloads/papers/019-WCSP2016a/ILGNet-Final.pdf

@chullhwan-song
Copy link
Owner Author

chullhwan-song commented Feb 25, 2019

Abstract

  • aesthetic quality에 대해 low or high로 분류하려는 문제. = image aesthetics classification problem
    • quite a challenging problem beyond image recognition.
  • 이 연구에서는 이러한 문제를 풀기위한 ILGNet라는 새로운 network 제시
    • inception module 응용. > googlenet
    • 중간에, local layer 와 global layer의 connect ?
    • imagenet pre-trained model
    • ava dataset
      image

Image Aesthetic Quality Assessment challenges

  • intra class네에서의 high or low 차이점 감지.
  • high와 low에 대한 aesthetics rules
  • Aesthetic 평가에 대한 인간의 주관적 평가 문제.

IMAGE AESTHETICS CLASSIFICATION VIA ILGNET

  • Inception Module
    image
    • max-pooling layers with stride 2 > 1
  • The ILGNet for Aesthetics Prediction
    image
    • 첫번째와 두번째는 local layer+ adding 256 fc
    • 마지막 last inception layer는 global layer + adding 512 fc
    • 최종적으로 concat = 256*2+512 -> fc layer
    • ILGNet에대한 imagenet training - for finetuning

실험

image
image

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

No branches or pull requests

1 participant