tubbz-alt / learn_dl_together

learn_DL_together 個人ノート

JTPA みんなでやろうDL オンライン勉強会 https://github.com/JTPA/learn_DL_together

Nocnoc: https://nocnoc.ooo/event/A1B6CDAC-637F-4455-9582-D086AC289268

何か自分でまとめないと学習しないのでGithubに載せてみました。まだ初心者の練習レベルなのでこのページ自体はシェアはしないで下さい。（以下に引用したリンクはご自由にシェアして下さい。）

オーディオ認識

もともと画像認識の方が興味があるのですが、写真技術のチュートリアルビデオ（MLとは関係なし）を見ていたら、シャッターの音がする度に講師が撮影した写真が画面に出るというビデオがあって、シャッター音を認識して画像フレームをセーブできないかというところからオーディオ認識の勉強を始めました。シャッター音のタイムスタンプがわかれば、そこの画像フレームをビデオから抽出するのは簡単です。

音声のスペクトル (MFCC) を取って画像のようにCNNをかければいいみたいです。以下のサイトはスピーチ音声等の認識ですが、シャッター音検出のために他にもっと良い方法があれば教えて下さい。

見つけたサイト

• https://medium.com/manash-en-blog/building-a-dead-simple-word-recognition-engine-using-convnet-in-keras-25e72c19c12b Building a Dead Simple Speech Recognition Engine using ConvNet in Keras by Manash Kumar Mandal
  • Code: https://github.com/manashmandal/DeadSimpleSpeechRecognizer

Colabに持ってきて実行 https://colab.research.google.com/drive/1WEFVUwM76hgqMMvHFDU_73hlWjQP3RQG?usp=sharing

• https://medium.com/@mikesmales/sound-classification-using-deep-learning-8bc2aa1990b7 Sound Classification using Deep Learning by Mike Smales

  • Code: https://github.com/mikesmales/Udacity-ML-Capstone 今のところエラーが起きてうまく動かない（詳細後日）

画像認識

• https://colab.research.google.com/github/google/automl/blob/master/efficientdet/tutorial.ipynb#scrollTo=V8-yl-s-WKMG EfficientDet Tutorial: inference, eval, and training

(https://github.com/JTPA/learn_DL_together のREADME.mdの中にリンクあり）

My copy: https://colab.research.google.com/drive/1O9zETdRbCL-HlfwHtSrRl7QotiBqMPI-#scrollTo=fHU46tfckaZo

inference.pyの最後の関数inferenceに犬とかバイク・車の絵をモデルに喰わせると、Bounding Boxやラベルの配列を返してくれる模様。 utils.pyのget_feat_sizesやanchors.pyのコードで、元画像から色々な大きさの部分画像を色々な位置から抽出して、それをモデルが犬、車、バイクと判断するかどうか見ている？

2020-07-01 DL／機械学習 オンライン勉強会 「みんなでやろうDL」 Day 4.6 もくもく会

https://www.meetup.com/JTPA-Japanese-Technology-Professionals-Association/events/271555823/

https://nocnoc.ooo/app#/chat/A1B6CDAC-637F-4455-9582-D086AC289268

(敬称略)

• https://ai.facebook.com/blog/end-to-end-object-detection-with-transformers/
  End-to-end object detection with Transformers
  Nicolas Carion, Sergey Zagoruyko, Francisco Massa - Facebook AI
  from Yuki
  

  • https://www.youtube.com/watch?v=Uumd2zOOz60
    How I Read a Paper: Facebook's DETR (Video Tutorial)
    Yannic Kilcher
    from Yuki

• https://openaccess.thecvf.com/content_CVPR_2020/html/Menon_PULSE_Self-Supervised_Photo_Upsampling_via_Latent_Space_Exploration_of_Generative_CVPR_2020_paper.html
  PULSE: Self-Supervised Photo Upsampling via Latent Space Exploration of Generative Models
  Sachit Menon, Alexandru Damian, Shijia Hu, Nikhil Ravi, Cynthia Rudin; The IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 2437-2445
  from Jin

• https://medium.com/@mikesmales/sound-classification-using-deep-learning-8bc2aa1990b7
  Sound Classification using Deep Learning
  Mike Smales
  

• https://medium.com/manash-en-blog/building-a-dead-simple-word-recognition-engine-using-convnet-in-keras-25e72c19c12b
  Building a Dead Simple Speech Recognition Engine using ConvNet in Keras
  Manash Kumar Mandal
  Code: https://github.com/manashmandal/DeadSimpleSpeechRecognizer
  Copied to Colab: https://colab.research.google.com/drive/1WEFVUwM76hgqMMvHFDU_73hlWjQP3RQG?usp=sharing
  

  • http://practicalcryptography.com/miscellaneous/machine-learning/guide-mel-frequency-cepstral-coefficients-mfccs/
    Mel Frequency Cepstral Coefficient (MFCC) tutorial
    
  • https://en.wikipedia.org/wiki/Mel-frequency_cepstrum
    Wikipedia: Mel-frequency cepstrum
    from Komei https://qiita.com/tmtakashi_dist/items/eecb705ea48260db0b62
    MFCC（メル周波数ケプストラム係数）入門
    Qiita: @tmtakashi_dist
    from Charlie Yoshida
  • https://www.tensorflow.org/api_docs/python/tf/signal/fft
    TensorFlow Core v2.2.0 tf.signal.fft
    from Jin

• https://arxiv.org/abs/1911.06971
  BSP-Net: Generating Compact Meshes via Binary Space Partitioning
  Zhiqin Chen, Andrea Tagliasacchi, Hao Zhang
  from oba
  

  • https://www.youtube.com/watch?v=9-ixexpjN-8
    BSP-Net: Generating Compact Meshes via Binary Space Partitioning, CVPR 2020, oral presentation video
    Zhiqin Chen
    
  • https://www.pyimagesearch.com/2016/11/07/intersection-over-union-iou-for-object-detection/
    Intersection over Union (IoU) for object detection
    Adrian Rosebrock - pyimagesearch.com
    from Komei
    
  • https://towardsdatascience.com/non-maximum-suppression-nms-93ce178e177c
    Non-maximum Suppression (NMS)
    Sambasivarao. K from Komei

• https://arxiv.org/abs/2003.08934
  NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis
  Ben Mildenhall, Pratul P. Srinivasan, Matthew Tancik, Jonathan T. Barron, Ravi Ramamoorthi, Ren Ng
  from サンディエゴの松原

• (https://arxiv.org/abs/2006.10739
  Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domains
  Matthew Tancik, Pratul P. Srinivasan, Ben Mildenhall, Sara Fridovich-Keil, Nithin Raghavan, Utkarsh Singhal, Ravi Ramamoorthi, Jonathan T. Barron, Ren Ng
  from サンディエゴの松原)

• https://arxiv.org/abs/1611.07004
  Image-to-Image Translation with Conditional Adversarial Networks
  Phillip Isola, Jun-Yan Zhu, Tinghui Zhou, Alexei A. Efros
  from サンディエゴの松原

  • https://www.youtube.com/watch?v=B1bMMF8miN8&pbjreload=101
    Image-to-Image Translation with Conditional Adversarial Networks // Creative AI Podcast Episode #11
    Creative AI Podcast
    from Komei

• https://www.renom.jp/ja/notebooks/tutorial/image_processing/yolo/notebook.html
  オブジェクト検出 YOLO
  from Charlie Yoshida

• https://qiita.com/mshinoda88/items/9770ee671ea27f2c81a9
  物体検出についての歴史まとめ(1)
  from Charlie Yoshida

• https://towardsdatascience.com/yolo-v5-is-here-b668ce2a4908
  YOLOv5 is Here!
  Mihir Rajput
  from DLまるでわからん

https://www.vox.com/recode/2020/6/29/21303588/deepfakes-anonymous-artificial-intelligence-welcome-to-chechnya How deepfakes could actually do some good from Komei

https://kantocv.connpass.com/event/178126/ 第三回　全日本コンピュータビジョン勉強会（前編） from Jin

• https://www.youtube.com/watch?v=9DdzzVmLzog JapanCV(7/4) from Jin

2020-07-15 DL／機械学習 オンライン勉強会 「みんなでやろうDL」 Day 5 もくもく会

https://www.meetup.com/JTPA-Japanese-Technology-Professionals-Association/events/271892957/

https://nocnoc.ooo/app#/chat/A1B6CDAC-637F-4455-9582-D086AC289268

• https://colab.research.google.com/drive/1vgrncRo-TwmoJu6JNXZiJGfZeWnb1vvJ?usp=sharing AutoKeras のTutorial (MNIST手書き数字の例）
• https://colab.research.google.com/drive/1_X4fz__f_nzSe1Qv273Hktr0gVeJra1R?usp=sharing AutokerasのPretainにてで物体認識 from Takeo Shiata / Jin

https://en.wikipedia.org/wiki/Elo_rating_system Elo rating system from Komei

https://medium.com/@gtnjuvin/my-journey-into-deep-q-learning-with-keras-and-gym-3e779cc12762 https://colab.sandbox.google.com/drive/13nOusO6ecKdql0dhNcuQ65jxcPNdNkDr https://github.com/GaetanJUVIN/Deep_QLearning_CartPole from Jin Copied to https://colab.research.google.com/drive/147br3HYVRgJYBGl0o2zhZ4tALUluhpkb#scrollTo=TUYqaYnvjkDN