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 |
- cv2
- ai공모전
- 탑싯
- 논문 pdf
- 토익문제
- scaico
- 토익공부
- 토익공부법
- 코드프렌즈
- 데이터분석
- pip
- 환급형프론트챌린지
- ADsP
- pdf 다운로드
- TOPCIT
- 미네르바 대학
- 데이터관련자격증
- arXiv
- ai consensus
- model collapse
- ai model collapse
- 탑싯시험
- minerva university
- toeic
- 토익문법
- 논문 pdf 이름
- students@ai seoul hackathon
- 토익
- 크롬 확장프로그램
- 프론트
- Today
- Total
목록ML DL (18)
토리의 데굴데굴 공부일기

3D scene generation task를 요즘 공부하고 있는데, 3D를 표현하는 표현 방식에도 아주 많은 방법들이 있다는 사실 다들 알고 계셨나요..? (저는 일단 잘 몰랐음) 그래서 이 기회에 제대로 한번 정리해보겠습니다! 1. 3D Scene Generation task 소개먼저 3D Scene generation task에 대해 알아봅시다~!이런 3D scene을 만드려면 어떤 기술을 사용해야 할까요?저는 처음에는 우리가 2D image를 만드는 것과 동일한 방식을 사용하면 되지 않을까 라고 막연하게 생각했어요. 하지만 막상 공부해보니 생각보다 훨씬 복잡한 기술이 얽혀 있었습니다. 특히 어떤 3D Representation(3D 표현방식)을 사용하는지에 따라 모델 구조, 훈련 방법도 완전히 달..

Paper: https://arxiv.org/abs/2504.13072Project page: https://zju3dv.github.io/hiscene/ HiScene: Creating Hierarchical 3D Scenes with Isometric View Generation.Scene-level 3D generation represents a critical frontier in multimedia and computer graphics, yet existing approaches either suffer from limited object categories or lack editing flexibility for interactive applications. In this paper, we ..

https://www.nature.com/articles/s41586-024-07566-y 네이쳐지에 실린 논문. Model Collapse의 개념을 새롭게 제안한 논문이다. LLM과 같은 GenAI 분야는 뜨거운 감자로 자리잡으며 굉장히 많은 연구가 진행되고 있다. 많은 사람들은 이제 일상생활 속에서도 ChatGPT와 같은 LLM을 활용해 작문에 도움을 받는다. 그럼 인터넷에는 점점 더 많은 GenAI가 생성한 인공 데이터가 쌓일 것이다. LLM들은 계속 인터넷 상의 데이터들을 활용해 학습한다고 하면, 그들은 자신들의 조상 GenAI가 출력한 데이터를 다시 학습하게 되는 것이다. 이 논문에서는 이러한 연속적인 과정이 Model Collapse를 야기할 수 있다고 말한다. 이것은 왜 문제가 되는 것..

https://www.arxiv.org/abs/2408.06292 The AI Scientist: Towards Fully Automated Open-Ended Scientific DiscoveryOne of the grand challenges of artificial general intelligence is developing agents capable of conducting scientific research and discovering new knowledge. While frontier models have already been used as aides to human scientists, e.g. for brainstorming iarxiv.org https://github.com/Sak..
원본 논문은 DBPia에서 찾으실 수 있습니다. 0. 요약기존 방법론실험 프로토콜을 의사코드로 변환하여 생성 결과를 평가하는 방식 → 의사코드 변환 방법 및 평가 한계단편적인 예시 기반 변환 방법본 연구파라미터 중심 변환 방법 → 프로토콜의 구조와 의미를 더욱 정확하게 반영유사도 기반 성능 평가 제시 → 연구 결과를 객관적으로 평가 가능1. 서론프로토콜: 실험의 재료, 과정, 분석 방법 등을 정리한 문서본 논문은 생명과학 분야의 실험 프로토콜 문서 자동 생성에 집중함기존 방법론대규모 언어 모델을 활용하여 자연어로 작성된 프로토콜을 의사코드 형식으로 변환하는 방법론 제시이 방법에서 one-shot 프롬프트 사용→ 의사코드 변환 과정에 대한 검증이 부족하다는 한계점본 연구의사코드화 프롬프트를 수정하는 새로..

Paperhttps://arxiv.org/abs/1801.07698 ArcFace: Additive Angular Margin Loss for Deep Face RecognitionRecently, a popular line of research in face recognition is adopting margins in the well-established softmax loss function to maximize class separability. In this paper, we first introduce an Additive Angular Margin Loss (ArcFace), which not only has a clearxiv.org Official githubhttps://github.c..

Paperhttps://arxiv.org/abs/2303.07284 Align and Attend: Multimodal Summarization with Dual Contrastive LossesThe goal of multimodal summarization is to extract the most important information from different modalities to form output summaries. Unlike the unimodal summarization, the multimodal summarization task explicitly leverages cross-modal information to helparxiv.org Official Githubhttps://g..

https://arxiv.org/pdf/2010.11929 IntroductionNLP domain에서 시작된 Transformer를 CV domain으로 적용해 처음으로 큰 성과를 얻은 논문이 ViT이다. Transformer 아키텍쳐가 NLP domain에서는 거의 표준처럼 사용이 되고 있었지만, 이를 CV domain에 적용시키는 것은 한계가 있었다.CV는 원래 CNN이 지배적인 분야였기 때문에 CNN과 self-attention을 결합하려는 다양한 시도들이 이루어졌지만, scalability를 가지진 못했다.**scalability: 모델의 사이즈(용량)이 커질수록 성능도 그에 비례해 높아지는 특성 하지만 ViT(Vision Transformer)는 CNN을 사용하지 않고 transform..