Be the only one, not the best one

#KcBERT #Dataset #Corpus

안녕하세요, KcBERT 학습에 사용한 데이터셋을 Kaggle을 통해 공개합니다!

KcBERT는 네이버 뉴스 댓글 데이터 2019.01.01~2020.06.11자 '랭킹뉴스'의 댓글로 학습한 Pretrain BERT 모델이고, 이번에 공개한 Cleaned 데이터셋으로 학습을 진행했습니다.

데이터셋은 약 12GB의 댓글로 이뤄져 있습니다.

아래 캐글 링크에서 다운 받으시고 여러분만의 Pretrain을 진행해 보세요!

https://www.kaggle.com/junbumlee/kcbert-pretraining-corpus-korean-news-comments

안녕하세요, SLAM 공부하는 장형기입니다.

최근 Visual-SLAM 논문들이 새로 나오면서, 이전에 했던 강의 슬라이드 내용을 업데이트 하였습니다.

이전과 동일하게 Feature-based SLAM, Direct SLAM, Visual-Inertial SLAM, Deep Learning SLAM 내용이 있고, 추가로 SLAM의 배경 내용이 되는 Kalman filter / Monte-Carlo Localization, 2020년 신규 업데이트 된 논문들, 그리고 기술 소개에 대한 부분을 조금 더 추가하였습니다. 파일에 영상이 많아 슬라이드쉐어 보다 원본 파일을 보시는 것을 추천드립니다.

감사합니다.

비디오 링크 포함 ppt (추천):
[https://cv-learn.com/PPT-SLAM-63bfe2f3902840e1b9db8d5387050574](https://cv-learn.com/PPT-SLAM-63bfe2f3902840e1b9db8d5387050574)

저용량 ppt:
[https://www.slideshare.net/HyunggiChang/visualslam-in-1-day](https://www.slideshare.net/HyunggiChang/visualslam-in-1-day)

#돌리면서배우는SLAM

안녕하세요, lidar slam 공부중인 기섭입니다.

오늘은 2D lidar slam 튜토리얼을 하나 작성해보았습니다.

SLAM을 몰라도 해보실 수 있을거같습니다.

지난 2019 ROSCon 에서 공개된

SLAM-toolbox 라는 프로그램을 돌리는 예제 및 설명입니다.

https://www.notion.so/giseopkim/SLAM-toolbox-aac021ec21d24f898ce230c19def3b7b

감사합니다~

캐글 첫 4x 그랜드마스터인 Abhishek Thakur가 머신러닝 딥러닝 코스를 기획중인것 같습니다

상당히 다양한 주제를 커버할 것으로 보이는데요, 첨부파일이 코스 드래프트를 보여줍니다. 신청 양식은 아래 링크에 있구요 :)

https://docs.google.com/forms/d/e/1FAIpQLSfzU_W7KjzATgCtFMajCnUGO4XwsxBqebRgv1VeRGYwGUAsHw/viewform

https://github.com/rbhatia46/Data-Science-Interview-Resources

I open-sourced a comprehensive guide to prepare for DataScience/AI interviews. The GitHub repo has hit 400+ stars at full tilt. The traffic received per day is insane, and I commit to continue helping by adding more resources.

# 질문 있습니다!

첨부한 파일은 성별 전환 애플리케이션(Application)으로 레오나르도 디카프리오(Leonardo DiCaprio)의 성별을 '여성'으로 전환시킨 영상입니다.

요는 어떤 GAN 모델을 사용하면 위와 같은 작업을 할 수 있을까요?

StarGAN Version 2 논문 1쪽에 나오는 그림 1을 보고 판단해보면 StarGAN Version 2로도 가능할까요?

Deep Learning Basics
By Daniel Worrall, MLSS Indo2020
ML Basics Video: https://youtu.be/FrbWQDdGpHQ?t=40
Slides: https://deworrall92.github.io/docs/MLSSIndo1_lo_res.pdf
DL Basics Video: https://youtu.be/K59cmobQKew?t=270
Slides: https://deworrall92.github.io/docs/MLSSIndo2_lo_res.pdf
Equivariance Video: https://youtu.be/HPU--yAGIBQ?t=407
Slides: https://deworrall92.github.io/docs/MLSSIndo3_lo_res.pdf
#ArtificialIntelligence #DeepLearning #MachineLearning

Python Numpy Tutorial with Colab
CS231n Python Tutorial With Google Colab : https://colab.research.google.com/github/cs231n/cs231n.github.io/blob/master/python-colab.ipynb
#Python #Numpy #Tutorial #Colab #DeepLearning

요즘 화제가 되고 있는 #OpenAI의 #GPT3 원리를 멋진 애니메이션 비주얼과 함께 설명한 글입니다. (영어)

[https://jalammar.github.io/how-gpt3-works-visualizations-animations/](https://jalammar.github.io/how-gpt3-works-visualizations-animations/)

#schoolofai #openai #gpt3 #nlp #machinelearning

COCO-WholeBody dataset is the first large-scale benchmark for whole-body pose estimation. It is an extension of COCO 2017 dataset with the same train/val split as COCO.

For project and code/API/expert requests: https://www.catalyzex.com/paper/arxiv:2007.11858

For each person, they annotate 4 types of bounding boxes (person box, face box, left-hand box, and right-hand box) and 133 keypoints (17 for body, 6 for feet, 68 for face and 42 for hands).

Get the free ML code finder browser extension:
Chrome https://bit.ly/code_finder_chrome
Firefox https://bit.ly/code_finder_firefox.

Made With ML Topics
A collection of the best ML tutorials, toolkits and research organized by topic: https://madewithml.com/topics/
#DeepLearning #MachineLearning #Tutorials

https://www.facebook.com/groups/DeepNetGroup/permalink/1221456308247249/

Recommend a good book on statistics?

EDIT:
Ok, so to summarize for posterity:
Statistical inference by Casella
Openintro statistics
Statistics by David Freedman
All of statistics - wasserman
Applied statistics by Montgomery
And best of all :...The iron of statistic by Walid Miak

Also some courses (which I haven't checked myself):

https://www.takethiscourse.net/fundamentals-of-statistics/

https://www.edx.org/course/fundamentals-of-statistics And if you want the free MIT OCW version of this course, then search for the course 18.650 (Statistics for applications)

Thanks for the responses folks.
I think we can lock this now

• Arthur Chan

  

  You mean a "good book"? Consider to change your status text.

   

  • 좋아요

     

  •  · 

    답글 달기

  •  · 

    공유하기

  •  · 18주

   

•  

  Vikram Tomar

  He probably means exhaustive.

   

  • 좋아요

     

  •  · 

    답글 달기

  •  · 

    공유하기

  •  · 18주

   

•  

  Chandrashekhar Thejaswi

  Depends on which level, and the rigour you want..

  One good nook is

  "Statistical inference" by Casella

  

  4

   

  • 좋아요

     

  •  · 

    답글 달기

  •  · 

    공유하기

  •  · 18주

   

•  

  Walid Miak

  The iron of statistic 

  

  

   

  • 좋아요

     

  •  · 

    답글 달기

  •  · 

    공유하기

  •  · 18주

  답글 7개

•  

  Imran Abdul Ghani

  Reading a book seems to be difficult... better to go for an online course.

  

  

  3

   

  • 좋아요

     

  •  · 

    답글 달기

  •  · 

    공유하기

  •  · 18주

  답글 1개

•  

  Deep Maity

  Youtube

  

  1

   

  • 좋아요

     

  •  · 

    답글 달기

  •  · 

    공유하기

  •  · 18주
  •  · 

    수정됨

   

•  

  Aly Mostafa

  Openintro statistics

  

  1

   

  • 좋아요

     

  •  · 

    답글 달기

  •  · 

    공유하기

  •  · 18주

   

•  

  Randy Novak

  What’s my chance for finding the right book? 

  

   

  • 좋아요

     

  •  · 

    답글 달기

  •  · 

    공유하기

  •  · 18주

   

•  

  Ho Leung Ng

  For beginners, I like the text by David Freedman.

  

  2

   

  • 좋아요

     

  •  · 

    답글 달기

  •  · 

    공유하기

  •  · 18주
  •  · 

    수정됨

  답글 1개

•  

  Burhan Ahmad

  All of statistics - wasserman

   

  • 좋아요

     

  •  · 

    답글 달기

  •  · 

    공유하기

  •  · 18주

   

•  

  Adrian Markelov

  All of statistics by Wasserman is the ultimate book but as other have said follow a class while using it

   

  • 좋아요

     

  •  · 

    답글 달기

  •  · 

    공유하기

  •  · 18주

   

•  

  Sanket Mishra

  Applied statistics by Montgomery, however I preferred freedman and walpole....

   

  • 좋아요

     

  •  · 

    답글 달기

  •  · 

    공유하기

  •  · 18주

   

•  

  Karl Wiklund

  Depends on what you want to focus on:

  https://www.amazon.de/Probability-Random.../dp/0130200719...

   

  

  AMAZON.DE

  Probability and Random Processes With Applications to Signal Processing

  Probability and Random Processes With Applications to Signal Processing

   

  • 좋아요

     

  •  · 

    답글 달기

  •  · 

    공유하기

  •  · 18주

   

•  

  Harsha Nm

  Any Good online course on probability n statistics??

   

  • 좋아요

     

  •  · 

    답글 달기

  •  · 

    공유하기

  •  · 18주

   

•  

  Ifiok EffiongBenzene Asukwo

  I think any Schuam's series is ok

   

  • 좋아요

     

  •  · 

    답글 달기

  •  · 

    공유하기

  •  · 18주

Image Augmentation Is All You Need: Regularizing Deep Reinforcement Learning from Pixels
Kostrikov et al.: https://arxiv.org/abs/2004.13649
Code: https://github.com/denisyarats/drq
Website: https://sites.google.com/view/data-regularized-q
#DeepLearning #MachineLearning #ReinforcementLearning

COCO-WholeBody dataset is the first large-scale benchmark for whole-body pose estimation. It is an extension of COCO 2017 dataset with the same train/val split as COCO.

For project and code/API/expert requests: https://www.catalyzex.com/paper/arxiv:2007.11858

For each person, they annotate 4 types of bounding boxes (person box, face box, left-hand box, and right-hand box) and 133 keypoints (17 for body, 6 for feet, 68 for face and 42 for hands).

Get the free ML code finder browser extension:
Chrome https://bit.ly/code_finder_chrome
Firefox https://bit.ly/code_finder_firefox.

Latest from Stanford and Adobe Researchers: Inferring 3D human motion from video sequences that takes initial 2D and 3D pose estimates as input.
For project and code/expert/API requests: https://www.catalyzex.com/paper/arxiv:2007.11678.
This process produces motions that are significantly more realistic than those from purely kinematic methods, substantially improving quantitative measures of both kinematic and dynamic plausibility.

Python 오픈소스(Open Source) 분석 방법

https://zzsza.github.io/development/2020/07/19/opensource-analysis/?fbclid=IwAR352orsNjkEERq0g9X3ri1ogh2rfUCam8EdhLpWFOs_6l4b4nTibMtFxew

For the benefit of new folks in the field, here's a list of some well known and prestigious free online courses/books on Data Science and machine learning (for certificate you might have to pay though in some of these). Do add if you know any good free courses/resources in comments:

1) Machine Learning (Stanford University, Andrew Ng Course, Coursera)

This is a very respected course and widely accepted in machine learning industry.

https://www.coursera.org/learn/machine-learning

Do the assignments of this course in python as the course is in Octave language which isn't used that much in industry. Assignment solutions in Python.

https://github.com/dibgerge/ml-coursera-python-assignments

https://github.com/jdwittenauer/ipython-notebooks

2) Python Data Science Handbook (Jake VanderPlas)

Famous Python Data Science handbook topic-wise:

https://jakevdp.github.io/PythonDataScienceHandbook/

3) Fast.ai ML course (Jeremy Howard)

http://course18.fast.ai/ml

This course is designed by Kaggle 2 time competition winner Jeremy Howard and he is also chief scientist at Kaggle.

4) ML Course by IBM (edX)

https://www.edx.org/course/machine-learning-with-python-a-practical-introduct

5) Machine Learning Crash Course by Google

https://developers.google.com/machine-learning/crash-course/ml-intro

6) The Analytics Edge - Massachusetts Institute of Technology (MIT)

https://courses.edx.org/courses/course-v1:MITx+15.071x+1T2020/course/

7) Deep Learning (Andrew Ng, Coursera)

https://www.coursera.org/specializations/deep-learning

Latest from Max Planck researchers: State of the art in Shape and Pose Disentanglement for 3D Meshes!

For project and code/expert/API requests: https://www.catalyzex.com/paper/arxiv:2007.11341

The experiments on datasets of 3D humans, faces, hands and animals demonstrate the generality of our approach.

From #ECCV2020: Reconstruct a morphable shape, texture, and viewpoint from an image collection without 3D ground truth *and* 2D keypoints, allowing us to explore new categories like shoes!

For project and code/expert/API requests: https://www.catalyzex.com/paper/arxiv:2007.10982

Facebook released a Machine Learning algorithm (Multilevel Pixel Aligned Implicit Function For High Resolution 3D Human Digitization.) that can reconstruct or generate 3D pose by just looking at a single image. So, is this a replacement of Character Designers. What do you think guys?

Paper: https://arxiv.org/pdf/2004.00452.pdf
Code: https://github.com/facebookresearch/pifuhd
Colab: https://colab.research.google.com/drive/11z58bl3meSzo6kFqkahMa35G5jmh2Wgt

Credit: PIFuHD or Multilevel Pixel Aligned Implicit Function For Heigher Resolution 3D Human Digitization.

#MachineLearning #machinelearningalgorithms #facebookpost #AI #artificialintelligence #facebookページ

Deep Learning and Computer Vision Course List.

Computer Vision 3: Detection, Segmentation and Tracking - Technical University Munich
https://www.youtube.com/playlist?list=PLuv1FSpHurUd08wNo1FMd3eCUZXm8qexe

Stanford CS221: Artificial Intelligence: Principles and Techniques
https://www.youtube.com/playlist?list=PLuv1FSpHurUeIGea3o8H9oEsIFgw_mizO

TUM Lectures | Advanced Deep Learning
https://www.youtube.com/playlist?list=PLuv1FSpHurUcQi2CwFIVQelSFCzxphJqz

Reinforcement Learning by Deep Mind (Google) 2020
https://www.youtube.com/playlist?list=PLuv1FSpHurUe_hYTJz-cFH1zo_i6jL5JF

Statistical Machine Learning 2020: Lectures by Ulrike von Luxburg
https://www.youtube.com/playlist?list=PLuv1FSpHurUcSNZaGAkVsbwKRYKPd0sLF

Latest from Microsoft researchers: High-quality video inpainting!

For project and code/expert/API requests: https://www.catalyzex.com/paper/arxiv:2007.10247

They propose to learn a joint Spatial-Temporal Transformer Network (STTN) for video inpainting. Specifically, they simultaneously fill missing regions in all input frames by self-attention, and propose to optimize STTN by a spatial-temporal adversarial loss

안녕하세요.
제가 자연어처리 입문하면서 도움되었던 자료들 공유해보려고 합니다!
딥러닝을 이용한 자연어 처리 입문 [https://wikidocs.net/book/2155]
BERT 톺아보기
[http://docs.likejazz.com/bert/#fn:fn-2]
Dissecting BERT Part 1: The Encoder
[https://medium.com/dissecting-bert/dissecting-bert-part-1-d3c3d495cdb3]
LaRva 데뷰 2019 - 엄~청 큰 언어 모델 공장 가동기!
[https://deview.kr/2019/schedule/291]
ChrisMcCormickAI 유튜브 채널
[https://www.youtube.com/watch?v=FKlPCK1uFrc&list=PLam9sigHPGwOBuH4_4fr-XvDbe5uneaf6]
[https://www.youtube.com/watch?v=l8ZYCvgGu0o]
RoBERTa 논문
[https://arxiv.org/abs/1907.11692]
ALBERT 논문 / 논문 리뷰
[https://arxiv.org/abs/1909.11942]
[https://y-rok.github.io/nlp/2019/10/23/albert.html]
Huggingface/transformers 라이브러리[https://github.com/huggingface/transformers]
[https://huggingface.co/transformers/quicktour.html]
[https://huggingface.co/models]
한국어 Tokenizer 비교
[https://blog.pingpong.us/dialog-bert-tokenizer/]
TorchText 전처리 라이브러리 사용법
[https://wikidocs.net/65348]
GPT-2 설명 / BERT랑 비교
[http://jalammar.github.io/illustrated-gpt2/]
앞으로도 꾸준히 올려보겠습니다!

Latest from Baidu researchers: Automatic video inpainting algorithm that can remove traffic agents from videos and synthesize missing regions with the guidance of depth/point cloud!

For project and code/expert/API requests: https://www.catalyzex.com/paper/arxiv:2007.08854

In order to fill a target inpainting area in a frame, it is straightforward to transform pixels from other frames into the current one with correct occlusion.

#BERT #KcBERT

안녕하세요!

한국어 댓글 데이터셋으로 BERT Pretrain을 처음부터 진행해 만든 KcBERT를 공개합니다 :)

공개된 한국어 BERT는 대부분 한국어 위키, 뉴스 기사, 책 등 잘 정제된 데이터를 기반으로 학습한 모델입니다. 한편, 실제로 NSMC와 같은 댓글형 데이터셋은 정제되지 않았고 구어체 특징에 신조어가 많으며, 오탈자 등 공식적인 글쓰기에서 나타나지 않는 표현들이 빈번하게 등장합니다.

KcBERT는 위와 같은 특성의 데이터셋에 적용하기 위해, 네이버 뉴스에서 댓글과 대댓글을 수집해, 토크나이저와 BERT모델을 처음부터 학습한 Pretrained BERT 모델입니다.

KcBERT는 Huggingface의 Transformers 라이브러리를 통해 간편히 불러와 사용할 수 있습니다. (별도의 파일 다운로드가 필요하지 않습니다!)

좀더 자세한 내용은 아래 Github Repo를 참고해주세요! :D

[https://github.com/Beomi/KcBERT](https://github.com/Beomi/KcBERT)

EDA를 쉽게 해볼 수 있는 좋은 파이썬 라이브러리가 있습니다.

sweetviz
pandas-profiling

html 보고서도 만들어주고 그동안 고생하면서 한 것 생각하면 참 쉽고 좋아졌습니다.

모두 그 느낌을 받아보시길...

Agile Human Behavior Imitation by humanoid models!

For project and code/API/expert requests: https://www.catalyzex.com/paper/arxiv:2006.07364

This approach is the first humanoid control method that successfully learns from a large-scale human motion dataset (Human3.6M) and generates diverse long-term motions.

Better than Faceapp: State of the art in Facial Attribute Editing!

For project and code/API/expert requests: https://www.catalyzex.com/paper/arxiv:2007.05892

Existing approaches suffer from a serious compromise between correct attribute generation and preservation of the other information such as identity and background, because they edit the attributes in the imprecise area. To resolve this dilemma, we propose a progressive attention GAN (PA-GAN) for facial attribute editing.

Training and inference jupyter notebook on an ongoing kaggle competition "Global wheat detection"

Training code - https://github.com/Tessellate-Imaging/Monk_Object_Detection/tree/master/application_model_zoo (Example 18)

Kaggle inference kernel - https://www.kaggle.com/abhishek4273/starter-code-using-monk-object-detection-library

The task is challenging because of
📌 overlap of dense wheat plants
📌 wind blurring the photographs
📌 wheat heads getting morphed in yellow and green backgrounds

The competition will be active for next 20-25 days.

Happy Coding!!

[Github/Repo] Pytorch Metric Learning

딥러닝 모델을 훈련 시킨다는 것은 '어떤 Loss를 어떻게 줄일것이냐' 입니다. 물론 Loss를 줄인다고 그 모델이 좋은 모델일거란 보장은 없지만요.

그리고 이 loss를 유사도, 즉 원본과의 차이를 측정하여 모델을 훈련시키는 방법이 metric learning입니다. 원본과 얼마나 유사한지 distance(loss)를 측정하는 방식은 다양합니다.

그런 다양한 방법을 Pytorch에서 사용할 수 있게 만들어둔 레포가 있어 공유합니다. triplet loss를 제외하고는 다들 저에게 생소하네요. 여기 있는 내용을 공부하며 metric에 대해 좀 더 고민해봐야겠습니다.

https://github.com/KevinMusgrave/pytorch-metric-learning