æ©æ¢°å¦ç¿
TensorFlowã®RNNå®è£ ã¯ãµã³ãã«ãå°ãªãã ãã¤ãããã£ããµã³ãã«ã³ã¼ãã§ã¯ã éå®ãããä¸é¨ã®APIãã使ã£ã¦ããªããªã©å ¨ä½ãç¶²ç¾ ãã¥ããæããããã®ã§ã ãªãã¹ãå ¨ä½æãæãåºããããããã«ãèªåç¨ã«ã¡ã¢ã ï¼ã¨è¨ãå²ã«åºæ¬çãªAPIãã使ã£ã¦ãªãâ¦
PyTorchã®DataLoaderã®ãã°ã§GPUã¡ã¢ãªã解æ¾ãããªããã¨ãããï¼ nvidia-smiã§è¦ã¦ã該å½ããã»ã¹idã¯è¡¨ç¤ºãããªãï¼ ä¸ã®ã³ãã³ãã§ç¡çãã解æ¾ã§ããï¼ ps aux|grep <username>|grep python|awk '{print $2}'|xargs kill</username>
çµè«ã¨ãã¦ã¯ä»ã®ã¨ããä¸æããã£ã¦ããªãããã«è¦ããï¼ ä»å¾ã®é²å±ã«ã¨ã¦ãæå¾ ï¼ Audio texture synthesis and style transfer https://dmitryulyanov.github.io/audio-texture-synthesis-and-style-transfer/ ååºã¯ãªãã¨ããã°ã Gatysãã®ä¸çªæåã®â¦
Deep Image Prior https://arxiv.org/abs/1711.10925 ãç»åã¨ã¯ãããããã®ãã¨ããäºåæ å ±ãCNNã®æ§é èªä½ã«ããããåãã£ã¦ããã ã¨ããã¨ãããåºçºç¹ã¨ãã¦ã大éã®ç»åãã¼ã¿ã»ããã«ããå¦ç¿ãå¿ è¦ã¨ãããåä¸ç»åã®ã¿ç¨ãã¦å¦ç¿ãè¡ã£ã¦ãã®ç»â¦
conditional GANã®ã©ãã«ã®ä¸ãæ¹ã¯è²ã ããã æ¯åã©ãããã°è¯ããããããããè¿·ã£ã¦ãã¾ãã githubã®å®è£ ãã¿ãã¨æ§ã ã«æ¸ããã¦ããã æç®ãèªããããè²ããªäººã®å®è£ ãæ¼ãã»ããç¥è¦ã貯ã¾ããã®é ã ä»åã¯MNISTã«å¯¾ãã¦DRAGANãç¨ãã¦ã ãã®ä¸ã§â¦
Swish: a Self-Gated Activation Function https://arxiv.org/abs/1710.05941 ReLUã®ä»£ããã«ãªãæ´»æ§åé¢æ°Swishã®ææ¡ã Swishã®å½¢ã¯ã·ã³ãã«ã $$ f(x) = x \cdot Ï (x) $$ ReLUã¨ã®éãã¯ï¼non-monotonicityã¨smoothnessã ã¨æãï¼ å°ããè² ã®å ¥åã¯ReLUâ¦
Intriguing properties of neural networks https://arxiv.org/abs/1312.6199 adversarial exampleãåå¼·ãããã¦èªãã ã ã¡ã¢ã æ¬æã§ã¯DNNã®ç´è¦³çã§ãªã以ä¸ã®2ã¤ã®æ§è³ªã«ã¤ãã¦è§¦ãã¦ãã åå¥ã¦ãããã観å¯ããã®ããè¤æ°ã¦ãããã®çµåãã観å¯ããâ¦
MakeGirlsMoe㯠é©ç°çãªçæç»åã®è³ªã§ããçä¼ã¿ä¸ã®ä¸éãé©ãããã ãã®MakeGirlsMoeã® ãããã¯ã¼ã¯æ§é (SRResNet) ç®çé¢æ°(DRAGAN) ã使ã£ã¦ãããã¨ãç»åãçæãã¦ã¿ãã DRAGANã«ã¤ãã¦ã¯ã以åã«è¨äºãæ¸ãã¦ããã å½æãã¾ã注ç®ããã¦ããªãâ¦
æè¿ãæ¥åã®å¿ããã¨ã å¥ä»¶ã§é²è¡ãã¦ããã¢ãã«ã®ãã©ã¡ã¼ã¿ãã¥ã¼ãã³ã°ã®åæããªãããã ããã°æ´æ°ããã°ããè¡ã£ã¦ããªãã£ãã ä¸æ¦ãæ´æ°ãé絶ãã¦ãã¾ãã¨ããªããªãã¢ããã¼ã·ã§ã³çã«åéãã¥ããã®ã§ã 以åã試ãããã¨ã®ãã£ã¦ãæ軽ãªpixâ¦
Abnormal Event Detection in Videos using Generative Adversarial Nets https://arxiv.org/abs/1708.09644 ç°å¸¸æ¤ç¥ã®é£ãã æ¢åã®ç°å¸¸ãã¼ã¿ã»ããã®ãµã³ãã«ãµã¤ãºãå°ããã㨠ç°å¸¸ã®å®ç¾©ãã¯ã£ãããã¦ããªãã㨠ãããã®èª²é¡ã«å¯¾ãã¦ãæ£å¸¸ãªãã¿ã¼â¦
InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets https://arxiv.org/abs/1606.03657 æ師ãªãå¦ç¿ã§çæç»åãå¶å¾¡ã§ããããªinfoGANã試ãã¦ã¿ãã çæãã¦ã¿ãç»åããã¡ã èæ¯ GANã®å¦ç¿ã§ã¯G(â¦
Unsupervised Anomaly Detection with Generative Adversarial Networks to Guide Marker Discovery https://arxiv.org/abs/1703.05921 ã¾ãä¸æºåã¨ãã¦ï¼æ£å¸¸ãªç»åãè¨ç·´ãã¼ã¿ã¨ãã¦ä½¿ãï¼GANãå¦ç¿ãããï¼ å¦ç¿å¾ã®GANã® ã¯latent space representatioâ¦
Learning from Simulated and Unsupervised Images through Adversarial Training https://arxiv.org/abs/1612.07828 æ¦è¦ å¦ç¿ã®ããã®ç»åããªãå ´åãä¸è¶³ãã¦ããå ´åï¼ã·ãã¥ã¬ã¼ã¿çã使ã人工çã«ç»åãçæãããã¨ãããï¼ ãããã·ãã¥ã¬ã¼ã¿ã§çâ¦
Coarse-to-Fine Volumetric Prediction for Single-Image 3D Human Pose https://arxiv.org/abs/1611.07828 3次å é¢ç¯æ¨å®åé¡ã«å¯¾ãã¦ã volumetric heatmapãå帰ããã å復çãªãããã¯ã¼ã¯ã§æ®µéçã«ç²¾åº¦ãä¸ãã é¢é£ç 究 CNNã«ããä¸æ¬¡å 姿å¢æ¨å®ã§ãâ¦
Adversarial PoseNet: A Structure-aware Convolutional Network for Human Pose Estimation https://arxiv.org/abs/1705.00389 é¢é£ç 究 姿å¢æ¨å®ã§ã¯DCNN(Deep Convolutional Neural Nets)ã§heatmapãå帰ããææ³ãä¸è¬ç ãããã以ä¸ã®å ´åã§ã¯ç¾å®ã§ã¯â¦
Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields arXiv:https://arxiv.org/abs/1611.08050 æ¦è¦ CNNã®çµåãã§ç»åå ã®äººç©å§¿å¢ãæ¤ç¥ãéæ NP-hardåé¡ãrelaxationãè¨ãã¦è§£ã top-downã¢ããã¼ãï¼äººãæ¤ç¥âæ¤ç¥ãã人ããããã®â¦
Stacked Hourglass Networks for Human Pose Estimation https://arxiv.org/abs/1603.06937 æ¦è¦ stacked hourglassãªå½¢ç¶ã®ãããã¯ã¼ã¯ã使ã£ã¦å§¿å¢æ¨å®ãã ç»åã®å ¨ã¦ã®å¤§ããã®æ å ±ãæãã¦å©ç¨ã§ãã hourglassãé£çµãhourglassãã¨ã«æ師ãã¼ã¿ãä¸â¦
Convolutional Pose Machines https://arxiv.org/abs/1602.00134 姿å¢èªèã®ç 究ã®æµãã追ããããã¨ããããã¬ã³ã¬ã³èªãã§ããã ãã¨ã§ç¶ºéºã«ã¾ã¨ãããã é¢é£ç 究 pictorial structures é¨ä½éã®ç©ºéçé¢ä¿ãæ¨æ§é ã®ã°ã©ãã£ã«ã«ã¢ãã«ã§è¨è¿° åè¢ãâ¦
Efficient Object Localization Using Convolutional Networks https://arxiv.org/abs/1411.4280 é¢é£ç 究 DeepPose (DeepPoseè«æã¡ã¢ - ç·è¶æèããã°) Toshev et. al,ã¯"FLIC"ã"LSP"ã®ãã¼ã¿ã»ããã§SOTA é¢ç¯ä½ç½®ãå帰åé¡ã¨ãã¦ç´æ¥è§£ã RGBç»åããâ¦
DeepPose: Human Pose Estimation via Deep Neural Networks http://www.cv-foundation.org/openaccess/content_cvpr_2014/papers/Toshev_DeepPose_Human_Pose_2014_CVPR_paper.pdf é¢é£ç 究 人ä½ãlocalãªãã¼ããé£çµãããã®ã¨è¦ãææ³ Figure Drawing Piâ¦
cs231nã®ææ¥ã¬ãã¼ããèªãã ã http://cs231n.stanford.edu/reports/2016/pdfs/426_Report.pdf Li 2014ã®ã¢ãã«ãã»ã¼ãã®ã¾ã¾ä½¿ã£ã¦ãã ä¸è¨ã¨ã®éãç¹ã¯ å ¥åã3ãã£ã³ãã«ã§ã¯ãªãã1ãã£ã³ãã«ã§ããã㨠conv2ã®å¾ã®æ£è¦åãããªãã㨠åºåã¯åé¢ç¯â¦
Revisiting Unreasonable Effectiveness of Data in Deep Learning Era https://arxiv.org/abs/1707.02968 ãã¼ã¿ãå¢ããã¨ã©ããªãã (a)ã¢ãã«ãµã¤ãºã(b)è¨ç®åã¯åä¸ãã¦ãããã(cï¼ãã¼ã¿ã»ãããµã¤ãºã®å¤åã¯å°ããã ãã¼ã¿ã»ããã大ãããªãã°ãâ¦
DRAGAN arXiv:https://arxiv.org/abs/1705.07215 âHow to train your DRAGAN"ã¨ããã¿ã¤ãã«ã®è«æã§ã å¤ãªã¿ã¤ãã«ã ãªã..ã¨æã£ã¦ãããã ãã®ã¿ã¤ãã«ã®å ãã¿ã¨ãã¦ãã¢ã¡ãªã«ã®3DCGã¢ãã¡ãããã®ãç¥ã£ãã ï¼æ¥æ¬åã¯ããã¯ã¨ãã©ã´ã³ã¨ãããããâ¦
Do GANs actually learn the distribution? An empirical study https://arxiv.org/abs/1706.08224 GANã«ãã£ã¦çæããããã¼ã¿ã®åå¸ã®å¤æ§æ§ãè©ä¾¡ããè«æã ãã®è«æã§ã¯èªçæ¥ã®ãã©ããã¯ã¹ã使ã£ãå°ãããªããã¼ãªåæããã¦ããã å®éã®GANã®çæâ¦
Conditional Generative Adversarial Nets https://arxiv.org/abs/1411.1784 cGANã¯æ¡ä»¶ä»ã確çåå¸ãå¦ç¿ããGANã ã¹ã¿ã³ãã¼ããªGANã§ã¯ï¼æå®ã®ç»åãçæãããã¨ãã£ããã¨ãé£ããï¼ ä¾ãã°0,1,â¦9ã®æ°åãçæãããããå¦ç¿ãããGANã«å¯¾ãã¦ã¯ï¼ â¦
cGAN-based Manga Colorization Using a Single Training Image arXiv:https://arxiv.org/abs/1706.06918 æ¦è¦ å¦ç¿ãã¼ã¿ã¨ãã¦ãã³ã¬ãéããã®ãé£ãã æ®éãã³ã¬ã¯ç½é»ãããªã copyrightã®åé¡ èªåè²ã¤ãææ³ã¨ãã¦pix2pixããããï¼å¤§éã®è¨ç·´ãã¼ã¿â¦
DeepLearningBookã§èªãã maxoutã«ã¤ãã¦ã®ã¡ã¢ï¼ Maxoutã¯ReLUãä¸è¬åãããã®ï¼ Maxoutã¦ãããã¯åã®å¤ãããªãã¦ãããã®éåã§ããï¼ ä»ã®æ´»æ§åé¢æ°ã¨éã£ã¦maxoutã§ã¯ï¼å層ã¦ãããããmaxoutå ã®åã¦ãããã¸ã®ç·å½¢å¤æã®ãã©ã¡ã¼ã¿ãå¦ç¿ããï¼ â¦
Image-to-Image Translation with Conditional Adversarial Networks arXiv:https://arxiv.org/abs/1611.07004 project:https://phillipi.github.io/pix2pix/ ç½é»åç»ã®colorizationãpix2pixã使ã£ã¦è¡ã£ã¦ã¿ãã¨ããã®ãæ¬è¨äºã®ä¸»æ¨ pix2pixã®æ¦è¦ ç»åâ¦
Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift arXiv:https://arxiv.org/abs/1502.03167 Internal Covariance Shiftã®åé¡ ãã¥ã¼ã©ã«ãããã¯ã¼ã¯ã§ã¯ï¼å ¥åã¨ãªããã¼ã¿ã®åå¸ãç½è²åããã¦ããã¨å¦ç¿â¦
Residual Networks Behave Like Ensembles of Relatively Shallow Networks arXiv:https://arxiv.org/abs/1605.06431 ResNetã®ååºã®è«æãèªãã ãã ãªã深層ã®å¦ç¿ããã¾ãè¡ã£ãã®ãä¸æçã ã£ãã æ¬è«æã§ã¯ãã®ããªãï¼ãã®é¨åã«å¯¾ãã解éãä¸ãã¦â¦