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 |
- Diffusion
- Stat110
- cv
- GAN
- animation retargeting
- Vision Transformer
- 언어모델
- CNN
- inductive bias
- 디퓨전모델
- BERT
- 딥러닝
- NLP
- deep learning
- dl
- Generative Model
- SGD
- 생성모델
- WinAPI
- 언리얼엔진
- 폰트생성
- ddpm
- Font Generation
- multimodal
- WBP
- UE5
- VAE
- RNN
- Few-shot generation
- cs231n
- Today
- Total
목록폰트생성 (4)
Deeper Learning
![](http://i1.daumcdn.net/thumb/C150x150/?fname=https://blog.kakaocdn.net/dn/SFn1X/btrPfYfewvb/ctAIyc7KZla1dtaeuwrck1/img.png)
Yitian Liu, Zhouhui Lian, [Wangxuan Institute of Computer Technology, Peking University, Beijing, China] (2022.10.13) Abstract 적은 데이터로 고품질의 중국어 폰트를 생성해내는 것은 어려운 일이며 현존하는 few-shot 폰트 생성 방식은 low-resolution의 획이 끊기는 폰트를 만드는데 그쳤다 이러한 문제를 해결하기 위해 본 논문은 고품질의 few-shot 생성이 가능한 FontTransformer를 제시 key idea는 prediction error가 쌓이는 것을 피하기 위한 parallel Transformer, 생성된 획의 퀄리티를 높이기 위한 serial Transformer의 사용 실제 ..
![](http://i1.daumcdn.net/thumb/C150x150/?fname=https://blog.kakaocdn.net/dn/49wGp/btrG4fhznUw/wjpQOu9Q9vkDyIIISXGEgK/img.png)
Licheng Tang, Yiyang Cai, Jiaming Liu, Zhibin Hong, Mingming Gong, Minhu Fan, Junyu Han, Jingtuo Liu, Errui Ding, Jingdong Wang [Baidu Inc, University of California, Berkeley, University of Melbourne] (2022.05) Abstract **Few-shot font generation(FFG)**는 노동 cost를 크게 줄여주어 주목을 받고 있는 기술 이전의 연구들은 reference glyph의 content와 style을 global, component-wise하게 disentangle 하는 방식으로 FFG에 접근 하지만 glyph의 style은 ..
![](http://i1.daumcdn.net/thumb/C150x150/?fname=https://blog.kakaocdn.net/dn/72jDf/btrwK9Z0Q1Z/mNOTy1KRPMfn0t8kF8im4K/img.png)
Seiya Matsuda, Akisato Kimura, Seiichi Uchida, Kyushu Univ., NTT, (2022.03) Abstract impression labeling 된 dataset을 사용하여 specific impression fonts를 GAN으로 생성하는 것이 목표 Main difficulty font impression은 애매 특정 impression label이 없다고 그 폰트가 해당 impression에 해당하지 않는 것이 아님 (dataset이 불안정) Key Idea co-occurrence-based missing label estimate impression label space compressor MyFonts 데이터셋은 전문가 + 비전문가가 tagging 한 ..
![](http://i1.daumcdn.net/thumb/C150x150/?fname=https://blog.kakaocdn.net/dn/z1F5P/btrfWMsutnD/Aq18iveF0V1CSkWrnpKmD0/img.png)
Abstract 현존 methods는 대부분 supervised learning, 매우 많은 paired data가 필요 image-to-image translation은 style을 텍스쳐와 색깔로 정의 냉림 Feature Deformation Skip Connection (FDSC)를 제시 predict pairs of displacement map employs the predicted maps to apply deformable convolution to the low-level feature maps from content encoder style-invariant feature 표현을 학습하기 위해 content encoder에 3개의 deformable convolutional layers를..